>>Welcome back to The Cube's coverage of Supercomputing Conference 2022, otherwise known as SC 22 here in Dallas, Texas. This is day three of our coverage, the final day of coverage here on the exhibition floor. I'm Dave Nicholson, and I'm here with my co-host, tech journalist extraordinaire, Paul Gillum. How's it going, >>Paul? Hi, Dave. It's going good. >>And we have a wonderful guest with us this morning, Dr. Panda from the Ohio State University. Welcome Dr. Panda to the Cube. >>Thanks a lot. Thanks a lot to >>Paul. I know you're, you're chopping at >>The bit, you have incredible credentials, over 500 papers published. The, the impact that you've had on HPC is truly remarkable. But I wanted to talk to you specifically about a product project you've been working on for over 20 years now called mva, high Performance Computing platform that's used by more than 32 organ, 3,200 organizations across 90 countries. You've shepherded this from, its, its infancy. What is the vision for what MVA will be and and how is it a proof of concept that others can learn from? >>Yeah, Paul, that's a great question to start with. I mean, I, I started with this conference in 2001. That was the first time I came. It's very coincidental. If you remember the Finman Networking Technology, it was introduced in October of 2000. Okay. So in my group, we were working on NPI for Marinette Quadrics. Those are the old technology, if you can recollect when Finman was there, we were the very first one in the world to really jump in. Nobody knew how to use Infin van in an HPC system. So that's how the Happy Project was born. And in fact, in super computing 2002 on this exhibition floor in Baltimore, we had the first demonstration, the open source happy, actually is running on an eight node infinite van clusters, eight no zeros. And that was a big challenge. But now over the years, I means we have continuously worked with all infinite van vendors, MPI Forum. >>We are a member of the MPI Forum and also all other network interconnect. So we have steadily evolved this project over the last 21 years. I'm very proud of my team members working nonstop, continuously bringing not only performance, but scalability. If you see now INFIN event are being deployed in 8,000, 10,000 node clusters, and many of these clusters actually use our software, stack them rapid. So, so we have done a lot of, like our focuses, like we first do research because we are in academia. We come up with good designs, we publish, and in six to nine months, we actually bring it to the open source version and people can just download and then use it. And that's how currently it's been used by more than 3000 orange in 90 countries. And, but the interesting thing is happening, your second part of the question. Now, as you know, the field is moving into not just hvc, but ai, big data, and we have those support. This is where like we look at the vision for the next 20 years, we want to design this MPI library so that not only HPC but also all other workloads can take advantage of it. >>Oh, we have seen libraries that become a critical develop platform supporting ai, TensorFlow, and, and the pie torch and, and the emergence of, of, of some sort of default languages that are, that are driving the community. How, how important are these frameworks to the, the development of the progress making progress in the HPC world? >>Yeah, no, those are great. I mean, spite our stencil flow, I mean, those are the, the now the bread and butter of deep learning machine learning. Am I right? But the challenge is that people use these frameworks, but continuously models are becoming larger. You need very first turnaround time. So how do you train faster? How do you do influencing faster? So this is where HPC comes in and what exactly what we have done is actually we have linked floor fighters to our happy page because now you see the MPI library is running on a million core system. Now your fighters and tenor four clan also be scaled to to, to those number of, large number of course and gps. So we have actually done that kind of a tight coupling and that helps the research to really take advantage of hpc. >>So if, if a high school student is thinking in terms of interesting computer science, looking for a place, looking for a university, Ohio State University, bruns, world renowned, widely known, but talk about what that looks like from a day on a day to day basis in terms of the opportunity for undergrad and graduate students to participate in, in the kind of work that you do. What is, what does that look like? And is, and is that, and is that a good pitch to for, for people to consider the university? >>Yes. I mean, we continuously, from a university perspective, by the way, the Ohio State University is one of the largest single campus in, in us, one of the top three, top four. We have 65,000 students. Wow. It's one of the very largest campus. And especially within computer science where I am located, high performance computing is a very big focus. And we are one of the, again, the top schools all over the world for high performance computing. And we also have very strength in ai. So we always encourage, like the new students who like to really work on top of the art solutions, get exposed to the concepts, principles, and also practice. Okay. So, so we encourage those people that wish you can really bring you those kind of experience. And many of my past students, staff, they're all in top companies now, have become all big managers. >>How, how long, how long did you say you've been >>At 31 >>Years? 31 years. 31 years. So, so you, you've had people who weren't alive when you were already doing this stuff? That's correct. They then were born. Yes. They then grew up, yes. Went to university graduate school, and now they're on, >>Now they're in many top companies, national labs, all over the universities, all over the world. So they have been trained very well. Well, >>You've, you've touched a lot of lives, sir. >>Yes, thank you. Thank >>You. We've seen really a, a burgeoning of AI specific hardware emerge over the last five years or so. And, and architectures going beyond just CPUs and GPUs, but to Asics and f PGAs and, and accelerators, does this excite you? I mean, are there innovations that you're seeing in this area that you think have, have great promise? >>Yeah, there is a lot of promise. I think every time you see now supercomputing technology, you see there is sometime a big barrier comes barrier jump. Rather I'll say, new technology comes some disruptive technology, then you move to the next level. So that's what we are seeing now. A lot of these AI chips and AI systems are coming up, which takes you to the next level. But the bigger challenge is whether it is cost effective or not, can that be sustained longer? And this is where commodity technology comes in, which commodity technology tries to take you far longer. So we might see like all these likes, Gaudi, a lot of new chips are coming up, can they really bring down the cost? If that cost can be reduced, you will see a much more bigger push for AI solutions, which are cost effective. >>What, what about on the interconnect side of things, obvi, you, you, your, your start sort of coincided with the initial standards for Infin band, you know, Intel was very, very, was really big in that, in that architecture originally. Do you see interconnects like RDMA over converged ethernet playing a part in that sort of democratization or commoditization of things? Yes. Yes. What, what are your thoughts >>There for internet? No, this is a great thing. So, so we saw the infinite man coming. Of course, infinite Man is, commod is available. But then over the years people have been trying to see how those RDMA mechanisms can be used for ethernet. And then Rocky has been born. So Rocky has been also being deployed. But besides these, I mean now you talk about Slingshot, the gray slingshot, it is also an ethernet based systems. And a lot of those RMA principles are actually being used under the hood. Okay. So any modern networks you see, whether it is a Infin and Rocky Links art network, rock board network, you name any of these networks, they are using all the very latest principles. And of course everybody wants to make it commodity. And this is what you see on the, on the slow floor. Everybody's trying to compete against each other to give you the best performance with the lowest cost, and we'll see whoever wins over the years. >>Sort of a macroeconomic question, Japan, the US and China have been leapfrogging each other for a number of years in terms of the fastest supercomputer performance. How important do you think it is for the US to maintain leadership in this area? >>Big, big thing, significantly, right? We are saying that I think for the last five to seven years, I think we lost that lead. But now with the frontier being the number one, starting from the June ranking, I think we are getting that leadership back. And I think it is very critical not only for fundamental research, but for national security trying to really move the US to the leading edge. So I hope us will continue to lead the trend for the next few years until another new system comes out. >>And one of the gating factors, there is a shortage of people with data science skills. Obviously you're doing what you can at the university level. What do you think can change at the secondary school level to prepare students better to, for data science careers? >>Yeah, I mean that is also very important. I mean, we, we always call like a pipeline, you know, that means when PhD levels we are expecting like this even we want to students to get exposed to, to, to many of these concerts from the high school level. And, and things are actually changing. I mean, these days I see a lot of high school students, they, they know Python, how to program in Python, how to program in sea object oriented things. Even they're being exposed to AI at that level. So I think that is a very healthy sign. And in fact we, even from Ohio State side, we are always engaged with all this K to 12 in many different programs and then gradually trying to take them to the next level. And I think we need to accelerate also that in a very significant manner because we need those kind of a workforce. It is not just like a building a system number one, but how do we really utilize it? How do we utilize that science? How do we propagate that to the community? Then we need all these trained personal. So in fact in my group, we are also involved in a lot of cyber training activities for HPC professionals. So in fact, today there is a bar at 1 1 15 I, yeah, I think 1215 to one 15. We'll be talking more about that. >>About education. >>Yeah. Cyber training, how do we do for professionals? So we had a funding together with my co-pi, Dr. Karen Tom Cook from Ohio Super Center. We have a grant from NASA Science Foundation to really educate HPT professionals about cyber infrastructure and ai. Even though they work on some of these things, they don't have the complete knowledge. They don't get the time to, to learn. And the field is moving so fast. So this is how it has been. We got the initial funding, and in fact, the first time we advertised in 24 hours, we got 120 application, 24 hours. We couldn't even take all of them. So, so we are trying to offer that in multiple phases. So, so there is a big need for those kind of training sessions to take place. I also offer a lot of tutorials at all. Different conference. We had a high performance networking tutorial. Here we have a high performance deep learning tutorial, high performance, big data tutorial. So I've been offering tutorials at, even at this conference since 2001. Good. So, >>So in the last 31 years, the Ohio State University, as my friends remind me, it is properly >>Called, >>You've seen the world get a lot smaller. Yes. Because 31 years ago, Ohio, in this, you know, of roughly in the, in the middle of North America and the United States was not as connected as it was to everywhere else in the globe. So that's, that's pro that's, I i it kind of boggles the mind when you think of that progression over 31 years, but globally, and we talk about the world getting smaller, we're sort of in the thick of, of the celebratory seasons where, where many, many groups of people exchange gifts for varieties of reasons. If I were to offer you a holiday gift, that is the result of what AI can deliver the world. Yes. What would that be? What would, what would, what would the first thing be? This is, this is, this is like, it's, it's like the genie, but you only get one wish. >>I know, I know. >>So what would the first one be? >>Yeah, it's very hard to answer one way, but let me bring a little bit different context and I can answer this. I, I talked about the happy project and all, but recently last year actually we got awarded an S f I institute award. It's a 20 million award. I am the overall pi, but there are 14 universities involved. >>And who is that in that institute? >>What does that Oh, the I ici. C e. Okay. I cycle. You can just do I cycle.ai. Okay. And that lies with what exactly what you are trying to do, how to bring lot of AI for masses, democratizing ai. That's what is the overall goal of this, this institute, think of like a, we have three verticals we are working think of like one is digital agriculture. So I'll be, that will be my like the first ways. How do you take HPC and AI to agriculture the world as though we just crossed 8 billion people. Yeah, that's right. We need continuous food and food security. How do we grow food with the lowest cost and with the highest yield? >>Water >>Consumption. Water consumption. Can we minimize or minimize the water consumption or the fertilization? Don't do blindly. Technologies are out there. Like, let's say there is a weak field, A traditional farmer see that, yeah, there is some disease, they will just go and spray pesticides. It is not good for the environment. Now I can fly it drone, get images of the field in the real time, check it against the models, and then it'll tell that, okay, this part of the field has disease. One, this part of the field has disease. Two, I indicate to the, to the tractor or the sprayer saying, okay, spray only pesticide one, you have pesticide two here. That has a big impact. So this is what we are developing in that NSF A I institute I cycle ai. We also have, we have chosen two additional verticals. One is animal ecology, because that is very much related to wildlife conservation, climate change, how do you understand how the animals move? Can we learn from them? And then see how human beings need to act in future. And the third one is the food insecurity and logistics. Smart food distribution. So these are our three broad goals in that institute. How do we develop cyber infrastructure from below? Combining HP c AI security? We have, we have a large team, like as I said, there are 40 PIs there, 60 students. We are a hundred members team. We are working together. So, so that will be my wish. How do we really democratize ai? >>Fantastic. I think that's a great place to wrap the conversation here On day three at Supercomputing conference 2022 on the cube, it was an honor, Dr. Panda working tirelessly at the Ohio State University with his team for 31 years toiling in the field of computer science and the end result, improving the lives of everyone on Earth. That's not a stretch. If you're in high school thinking about a career in computer science, keep that in mind. It isn't just about the bits and the bobs and the speeds and the feeds. It's about serving humanity. Maybe, maybe a little, little, little too profound a statement, I would argue not even close. I'm Dave Nicholson with the Queue, with my cohost Paul Gillin. Thank you again, Dr. Panda. Stay tuned for more coverage from the Cube at Super Compute 2022 coming up shortly. >>Thanks a lot.

(upbeat music) >> Welcome back. We're here at Supercomputing 22 in Dallas. My name's Paul Gill, I'm your host. With me, Dave Nicholson, my co-host. And one thing that struck me about this conference arriving here, was the number of universities that are exhibiting here. I mean, big, big exhibits from universities. Never seen that at a conference before. And one of those universities is Indiana University. Our two guests, Dave Jent, who's the AVP of Networks at Indiana University, Aaron Neal, Deputy CIO at Indiana University. Welcome, thanks for joining us. >> Thank you for having us. >> Thank you. >> I've always thought that the CIO job at a university has got to be the toughest CIO job there is, because you're managing this sprawling network, people are doing all kinds of different things on it. You've got to secure it. You've got to make it performant. And it just seems to be a big challenge. Talk about the network at Indiana University and what you have done particularly since the pandemic, how that has affected the architecture of your network. And what you do to maintain the levels of performance and security that you need. >> On the network side one of the things we've done is, kept in close contact with what the incoming students are looking for. It's a different environment than it was then 10 years ago when a student would come, maybe they had a phone, maybe they had one laptop. Today they're coming with multiple phones, multiple laptops, gaming devices. And the expectation that they have to come on a campus and plug all that stuff in causes lots of problems for us, in managing just the security aspect of it, the capacity, the IP space required to manage six, seven devices per student when you have 35,000 students on campus, has always been a challenge. And keeping ahead of that knowing what students are going to come in with, has been interesting. During the pandemic the campus was closed for a bit of time. What we found was our biggest challenge was keeping up with the number of people who wanted to VPN to campus. We had to buy additional VPN licenses so they could do their work, authenticate to the network. We doubled, maybe even tripled our our VPN license count. And that has settled down now that we're back on campus. But again, they came back with a vengeance. More gaming devices, more things to be connected, and into an environment that was a couple years old, that we hadn't done much with. We had gone through a pretty good size network deployment of new hardware to try to get ready for them. And it's worked well, but it's always challenging to keep up with students. >> Aaron, I want to ask you about security because that really is one of your key areas of focus. And you're collaborating with counties, local municipalities, as well as other educational institutions. How's your security strategy evolving in light of some of the vulnerabilities of VPNs that became obvious during the pandemic, and this kind of perfusion of new devices that that Dave was talking about? >> Yeah, so one of the things that we we did several years ago was establish what we call OmniSOC, which is a shared security operations center in collaboration with other institutions as well as research centers across the United States and in Indiana. And really what that is, is we took the lessons that we've learned and the capabilities that we've had within the institution and looked to partner with those key institutions to bring that data in-house, utilize our staff such that we can look for security threats and share that information across the the other institutions so that we can give each of those areas a heads up and work with those institutions to address any kind of vulnerabilities that might be out there. One of the other things that you mentioned is, we're partnering with Purdue in the Indiana Office of Technology on a grant to actually work with municipalities, county governments, to really assess their posture as it relates to security in those areas. It's a great opportunity for us to work together as institutions as well as work with the state in general to increase our posture as it relates to security. >> Dave, what brings IU to Supercomputing 2022? >> We've been here for a long time. And I think one of the things that we're always interested in is, what's next? What's new? There's so many, there's network vendors, software vendors, hardware vendors, high performance computing suppliers. What is out there that we're interested in? IU runs a large Cray system in Indiana called Big Red 200. And with any system you procure it, you get it running, you operate it, and your next goal is to upgrade it. And what's out there that we might be interested? That I think why we come to IU. We also like to showcase what we do at IU. If you come by the booth you'll see the OmniSOC, there's some video on that. The GlobalNOC, which I manage, which supports a lot of the RNE institutions in the country. We talk about that. Being able to have a place for people to come and see us. If you stand by the booth long enough people come and find you, and want to talk about a project they have, or a collaboration they'd like to partner with. We had a guy come by a while ago wanting a job. Those are all good things having a big booth can do for you. >> Well, so on that subject, in each of your areas of expertise and your purview are you kind of interleaved with the academic side of things on campus? Do you include students? I mean, I would think it would be a great source of cheap labor for you at least. Or is there kind of a wall between what you guys are responsible for and what students? >> Absolutely we try to support faculty and students as much as we can. And just to go back a little bit on the OmniSOC discussion. One of the things that we provide is internships for each of the universities that we work with. They have to sponsor at least three students every year and make that financial commitment. We bring them on site for three weeks. They learn us alongside the other analysts, information security analysts and work in a real world environment and gain those skills to be able to go back to their institutions and do an additional work there. So it's a great program for us to work with students. I think the other thing that we do is we provide obviously the infrastructure that enable our faculty members to do the research that they need to do. Whether that's through Big Red 200, our Supercomputer or just kind of the everyday infrastructure that allows them to do what they need to do. We have an environment on premise called our Intelligent Infrastructure, that we provide managed access to hardware and storage resources in a way that we know it's secure and they can utilize that environment to do virtually anything that they need in a server environment. >> Dave, I want to get back to the GigaPOP, which you mentioned earlier you're the managing director of the Indiana GigaPOP. What exactly is it? >> Well, the GigaPOP and there are a number of GigaPOP around the country. It was really the aggregation facility for Indiana and all of the universities in Indiana to connect to outside resources. GigaPOP has connections to internet too, the commodity internet, Esnet, the Big Ten or the BTAA a network in Chicago. It's a way for all universities in Indiana to connect to a single source to allow them to connect nationally to research organizations. >> And what are the benefits of having this collaboration of university. >> If you could think of a researcher at Indiana wants to do something with a researcher in Wisconsin, they both connect to their research networks in Wisconsin and Indiana, and they have essentially direct connection. There's no commodity internet, there's no throttling of of capacity. Both networks and the interconnects because we use internet too, are essentially UNT throttled access for the researchers to do anything they need to do. It's secure, it's fast, easy to use, in fact, so easy they don't even know that they're using it. It just we manage the networks and organize the networks in a way configure them that's the path of least resistance and that's the path traffic will take. And that's nationally. There are lots of these that are interconnected in various ways. I do want to get back to the labor point, just for a moment. (laughs) Because... >> You're here to claim you're not violating any labor laws. Is that what you're going to be? >> I'm here to hopefully hire, get more people to be interested to coming to IU. >> Stop by the booth. >> It's a great place to work. >> Exactly. >> We hire lots of interns and in the network space hiring really experienced network engineers, really hard to do, hard to attract people. And these days when you can work from anywhere, you don't have to be any place to work for anybody. We try to attract as many students as we can. And really we're exposing 'em to an environment that exists in very few places. Tens of thousands of wireless access points, big fast networks, interconnections and national international networks. We support the Noah network which supports satellite systems and secure traffic. It really is a very unique experience and you can come to IU, spend lots of years there and never see the same thing twice. We think we have an environment that's really a good way for people to come out of college, graduate school, work for some number of years and hopefully stay at IU, but if not, leave and get a good job and talk well about IU. In fact, the wireless network today here at SC was installed and is managed by a person who manages our campus network wireless, James Dickerson. That's the kind of opportunity we can provide people at IU. >> Aaron, I'd like to ask, you hear a lot about everything moving to the cloud these days, but in the HPC world I don't think that move is happening as quickly as it is in some areas. In fact, there's a good argument some workloads should never move to the cloud. You're having to balance these decisions. Where are you on the thinking of what belongs in the data center and what belongs in the cloud? >> I think our approach has really been specific to what the needs are. As an institution, we've not pushed all our chips in on the cloud, whether it be for high performance computing or otherwise. It's really looking at what the specific need is and addressing it with the proper solution. We made an investment several years ago in a data center internally, and we're leveraging that through the intelligent infrastructure that I spoke about. But really it's addressing what the specific need is and finding the specific solution, rather than going all in in one direction or another. I dunno if Jet Stream is something that you would like to bring up as well. >> By having our own data center and having our own facilities we're able to compete for NSF grants and work on projects that provide shared resources for the research community. Just dream is a project that does that. Without a data center and without the ability to work on large projects, we don't have any of that. If you don't have that then you're dependent on someone else. We like to say that, what we are proud of is the people come to IU and ask us if they can partner on our projects. Without a data center and those resources we are the ones who have to go out and say can we partner on your project? We'd like to be the leaders of that in that space. >> I wanted to kind of double click on something you mentioned. Couple of things. Historically IU has been I'm sure closely associated with Chicago. You think of what are students thinking of doing when they graduate? Maybe they're going to go home, but the sort of center of gravity it's like Chicago. You mentioned talking about, especially post pandemic, the idea that you can live anywhere. Not everybody wants to live in Manhattan or Santa Clara. And of course, technology over decades has given us the ability to do things remotely and IU is plugged into the globe, doesn't matter where you are. But have you seen either during or post pandemic 'cause we're really in the early stages of this. Are you seeing that? Are you seeing people say, Hey, thinking about their family, where do I want to live? Where do I want to raise my family? I'm in academia and no, I don't want to live in Manhattan. Hey, we can go to IU and we're plugged into the globe. And then students in California we see this, there's some schools on the central coast where people loved living there when they were in college but there was no economic opportunity there. Are you seeing a shift, are basically houses in Bloomington becoming unaffordable because people are saying, you know what, I'm going to stay here. What does that look like? >> I mean, for our group there are a lot of people who do work from home, have chosen to stay in Bloomington. We have had some people who for various reasons want to leave. We want to retain them, so we allow them to work remotely. And that has turned into a tool for recruiting. The kid that graduates from Caltech. Doesn't want to stay in Caltech in California, we have an opportunity now he can move to wherever between here and there and we can hire him do work. We love to have people come to Indiana. We think it is a unique experience, Bloomington, Indianapolis are great places. But I think the reality is, we're not going to get everybody to come live, be a Hoosier, how do we get them to come and work at IU? In some ways disappointing when we don't have buildings full of people, but 40 paying Zoom or teams window, not kind the same thing. But I think this is what we're going to have to figure out, how do we make this kind of environment work. >> Last question here, give you a chance to put in a plug for Indiana University. For those those data scientists those researchers who may be open to working somewhere else, why would they come to Indiana University? What's different about what you do from what every other academic institution does, Aaron? >> Yeah, I think a lot of what we just talked about today in terms of from a network's perspective, that were plugged in globally. I think if you look beyond the networks I think there are tremendous opportunities for folks to come to Bloomington and experience some bleeding edge technology and to work with some very talented people. I've been amazed, I've been at IU for 20 years and as I look at our peers across higher ed, well, I don't want to say they're not doing as well I do want brag at how well we're doing in terms of organizationally addressing things like security in a centralized way that really puts us in a better position. We're just doing a lot of things that I think some of our peers are catching up to and have been catching up to over the last 10, 12 years. >> And I think to sure scale of IU goes unnoticed at times. IU has the largest medical school in the country. One of the largest nursing schools in the country. And people just kind of overlook some of that. Maybe we need to do a better job of talking about it. But for those who are aware there are a lot of opportunities in life sciences, healthcare, the social sciences. IU has the largest logistics program in the world. We teach more languages than anybody else in the world. The varying kinds of things you can get involved with at IU including networks, I think pretty unparalleled. >> Well, making the case for high performance computing in the Hoosier State. Aaron, Dave, thanks very much for joining you making a great case. >> Thank you. >> Thank you. >> We'll be back right after this short message. This is theCUBE. (upbeat music)

>> From around the globe, it's theCUBE, with digital coverage of Exascale Day. Made possible by Hewlett Packard Enterprise. >> Hey, welcome back everybody Jeff Frick here with theCUBE. We are celebrating Exascale Day. 10, 18, I think it's the second year of celebrating Exascale Day, and we're really excited to have our next guest and talk about kind of what this type of compute scale enables, and really look a little bit further down the road at some big issues, big problems and big opportunities that this is going to open up. And I'm really excited to get in this conversation with our next guest. He is Bryan Pijanowski the Professor of Landscape and Soundscape Ecology at Purdue University. Bryan, great to meet you. >> Great to be here. >> So, in getting ready for this conversation, I just watched your TED Talk, and I just loved one of the quotes. I actually got one of quote from it that's basically saying you are exploring the world through sound. I just would love to get a little deeper perspective on that, because that's such a unique way to think about things and you really dig into it and explain why this is such an important way to enjoy the world, to absorb the world and think about the world. >> Yeah, that's right Jeff. So the way I see it, sound is kind of like a universal variable. It exists all around us. And you can't even find a place on earth where there's no sound, where it's completely silent. Sound is a signal of something that's happening. And we can use that information in ways to allow us to understand the earth. Just thinking about all the different kinds of sounds that exist around us on a daily basis. I hear the birds, I hear the insects, but there's just a lot more than that. It's mammals and some cases, a lot of reptiles. And then when you begin thinking outside the biological system, you begin to hear rain, wind, thunder. And then there's the sounds that we make, sounds of traffic, the sounds of church bells. All of this is information, some of it's symbolic, some of it's telling me something about change. As an ecologist that's what I'm interested in, how is the earth changing? >> That's great and then you guys set up at Purdue, the Purdue Center for Global Soundscapes. Tell us a little bit about the mission and some of the work that you guys do. >> Well, our mission is really to use sound as a lens to study the earth, but to capture it in ways that are meaningful and to bring that back to the public to tell them a story about how the earth kind of exists. There's an incredible awe of nature that we all experience when we go out and listen into to the wild spaces of the earth. I've gone to the Eastern Steppes of Mongolian, I've climbed towers in the Paleotropics of Borneo and listened at night. And ask the question, how are these sounds different? And what is a grassland really supposed to sound like, without humans around? So we use that information and bring it back and analyze it as a means to understand how the earth is changing and really what the biological community is all about, and how things like climate change are altering our spaces, our wild spaces. I'm also interested in the role that people play and producing sound and also using sound. So getting back to Mongolia, we have a new NSF funded project where we're going to be studying herders and the ways in which they use sonic practices. They use a lot of sounds as information sources about how the environment is changing, but also how they relate back to place and to heritage a special sounds that resonate, the sounds of a river, for example, are the resonance patterns that they tune their throat to that pay homage to their parents that were born at the side of that river. There's these special connections that people have with place through sound. And so that's another thing that we're trying to do. In really simple terms, I want to go out and, what I call it sounds rather simple, record the earth-- >> Right. >> What does that mean? I want to go to every major biome and conduct a research study there. I want to know what does a grassland sound like? What is a coral reef sound like? A kelp forest and the oceans, a desert, and then capture that as baseline and use that information-- >> Yeah. >> For scientific purposes >> Now, there's so much to unpack there Bryan. First off is just kind of the foundational role that sound plays in our lives that you've outlined in great detail and you talked about it's the first sense that's really activated as we get consciousness, even before we're born right? We hear the sounds of our mother's heartbeat and her voice. And even the last sense that goes at the end a lot of times, in this really intimate relationship, as you just said, that the sounds represent in terms of our history. We don't have to look any further than a favorite song that can instantly transport you, almost like a time machine to a particular place in time. Very, very cool. Now, it's really interesting that what you're doing now is taking advantage of new technology and just kind of a new angle to capture sound in a way that we haven't done before. I think you said you have sound listening devices oftentimes in a single location for a year. You're not only capturing sound, the right sound is changes in air pressure, so that you're getting changes in air pressure, you're getting vibration, which is kind of a whole different level of data. And then to be able to collect that for a whole year and then start to try to figure out a baseline which is pretty simple to understand, but you're talking about this chorus. I love your phrase, a chorus, because that sound is made up of a bunch of individual inputs. And now trying to kind of go under the covers to figure out what is that baseline actually composed of. And you talk about a bunch of really interesting particular animals and species that combine to create this chorus that now you know is a baseline. How did you use to do that before? I think it's funny one of your research papers, you reach out to the great bird followers and bird listeners, 'cause as you said, that's the easiest way or the most prolific way for people to identify birds. So please help us in a crowdsource way try to identify all the pieces that make this beautiful chorus, that is the soundscape for a particular area. >> Right, yeah, that's right. It really does take a team of scientists and engineers and even folks in the social sciences and the humanities to really begin to put all of these pieces together. Experts in many fields are extremely valuable. They've got great ears because that's the tools that they use to go out and identify birds or insects or amphibians. What we don't have are generalists that go out and can tell you what everything sounds like. And I'll tell you that will probably never ever happen. That's just way too much, we have millions of species that exist on this planet. And we just don't have a specific catalog of what everything sounds like, it's just not possible or doable. So I need to go out and discover and bring those discoveries back that help us to understand nature and understand how the earth is changing. I can't wait for us to eventually develop that catalog. So we're trying to develop techniques and tools and approaches that allow us to develop this electronic catalog. Like you're saying this chorus, and it doesn't necessarily have to be a species specific chorus, it can be a chorus of all these different kind of sounds that we think relate back to this kind of animal or that kind of animal based upon the animals instrument-- >> Right, great. >> And this is the sound. >> Now again, you know, keep it to the exascale theme, right? You're collecting a lot of data and you mentioned in one of the pieces I've dug up, that your longest study in a single location is 17 years. You've got over 4 million recordings. And I think you said over 230 years if you wanted to listen to them all back to back. I mean, this is a huge, a big data problem in terms of the massive amount of data that you have and need to run through an analysis. >> Yeah, that's right. We're collecting 48,000 data points per second. So that's 48 kilohertz. And then so you multiply everything and then you have a sense of how many data points you actually have to put them all together. When you're listening to a sound file over 10 minutes, you have hundreds of sounds that exist in them. Oftentimes you just don't know what they are, but you can more or less put some kind of measure on all of them and then begin to summarize them over space and time and try to understand it from a perspective of really science. >> Right, right. And then I just love to get your take as you progress down this kind of identification road, we're all very familiar with copyright infringement hits on YouTube or social media or whatever, when it picks up on some sound and the technology is actually really sophisticated to pick up some of those sound signatures. But to your point, it's a lot easier to compare against the known and to search for that known. Then when you've got this kind of undefined chorus that said we do know that there can be great analysis done that we've seen AI and ML applied, especially in the surveillance side on the video-- >> Right. >> With video that it can actually do a lot of computation and a lot of extracting signal from the noise, if you will. As you look down the road on the compute side for the algorithms that you guys are trying to build with the human input of people that know what you're listening to, what kind of opportunities do you see and where are we on that journey where you can get more leverage out of some of these technology tools? >> Well, I think what we're doing right now is developing the methodological needs, kind of describe what it is we need to move into that new space, which is going to require these computational, that computational infrastructure. So, for example, we have a study right now where we're trying to identify certain kinds of mosquitoes (chuckling) a vector-borne mosquitoes, and our estimates is that we need about maybe 900 to 1200 specific recordings per species to be able to put it into something like a convolutional neural network to be able to extract out the information, and look at the patterns and data, to be able to say indeed this is the species that we're interested in. So what we're going to need and in the future here is really a lot of information that allow us to kind of train these neural networks and help us identify what's in the sound files. As you can imagine the computational infrastructure needed to do that for data storage and CPU, GPU is going to be truly amazing. >> Right, right. So I want to get your take on another topic. And again the basis of your research is really all bound around the biodiversity crisis right? That's from the kind of-- >> Yeah. >> The thing that's started it and now you're using sound as a way to measure baseline and talk about loss of species, reduced abundancies and rampant expansion of invasive species as part of your report. But I'd love to get your take on cities. And how do you think cities fit the future? Clearly, it's an efficient way to get a lot of people together. There's a huge migration of people-- >> Right. >> To cities, but one of your themes in your Ted Talk is reconnecting with nature-- >> Yeah. >> Because we're in cities, but there's this paradox right? Because you don't want people living in nature can be a little bit disruptive. So is it better to kind of get them all in a tip of a peninsula in San Francisco or-- >> Yeah. >> But then do they lose that connection that's so important. >> Yeah. >> I just love to get your take on cities and the impacts that they're have on your core research. >> Yeah, I mean, it truly is a paradox as you just described it. We're living in a concrete jungle surrounded by not a lot of nature, really, honestly, occasional bird species that tend to be fairly limited, selected for limited environments. So many people just don't get out into the wild. But visiting national parks certainly is one of those kinds of experience that people oftentimes have. But I'll just say that it's getting out there and truly listening and feeling this emotional feeling, psychological feeling that wraps around you, it's a solitude. It's just you and nature and there's just no one around. >> Right. >> And that's when it really truly sinks in, that you're a part of this place, this marvelous place called earth. And so there are very few people that have had that experience. And so as I've gone to some of these places, I say to myself I need to bring this back. I need to tell the story, tell the story of the awe of nature, because it truly is an amazing place. Even if you just close your eyes and listen. >> Right, right. >> And it, the dawn chorus in the morning in every place tells me so much about that place. It tells me about all the animals that exist there. The nighttime tells me so much too. As a scientist that's spent most of his career kind of going out and working during the day, there's so much happening at night. Matter of fact-- >> Right. >> There's more sounds at night than there were during the day. So there is a need for us to experience nature and we don't do that. And we're not aware of these crises that are happening all over the planet. I do go to places and I listen, and I can tell you I'm listening for things that I think should be there. You can listen and you can hear the gaps, the gaps and that in that chorus, and you think what should be there-- >> Right. >> And then why isn't it there? And that's where I really want to be able to dig deep into my sound files and start to explore that more fully. >> It's great, it's great, I mean, I just love the whole concept of, and you identified it in the moment you're in the tent, the thunderstorm came by, it's really just kind of changing your lens. It's really twisting your lens, changing your focus, because that sound is there, right? It's been there all along, it's just, do you tune it in or do you tune it out? Do you pay attention? Do not pay attention is an active process or a passive process and like-- >> Right. >> I love that perspective. And I want to shift gears a little bit, 'cause another big environmental thing, and you mentioned it quite frequently is feeding the world's growing population and feeding it-- >> Yeah. >> In an efficient way. And anytime you see kind of factory farming applied to a lot of things you wonder is it sustainable, and then all the issues that come from kind of single output production whether that's pigs or coffee or whatever and the susceptibility to disease and this and that. So I wonder if you could share your thoughts on, based on your research, what needs to change to successfully and without too much destruction feed this ever increasing population? >> Yeah, I mean, that's one of the grand challenges. I mean, society is facing so many at the moment. In the next 20 years or so, 30 years, we're going to add another 2 billion people to the planet, and how do we feed all of them? How do we feed them well and equitably across the globe? I don't know how to do that. But I'll tell you that our crops and the ecosystem that supports the food production needs the animals and the trees and the microbes for the ecosystem to function. We have many of our crops that are pollinated by birds and insects and other animals, seeds need to be dispersed. And so we need the rest of life to exist and thrive for us to thrive too. It's not an either, it's not them or us, it has to be all of us together on this planet working together. We have to find solutions. And again, it's me going out to some of these places and bringing it back and saying, you have to listen, you have to listen to these places-- >> Right. >> They're truly a marvelous. >> So I know most of your listening devices are in remote areas and not necessarily in urban areas, but I'm curious, do you have any in urban areas? And if so, how has that signature changed since COVID? I just got to ask, (Bryan chuckling) because we went to this-- >> Yeah. >> Light switch moment in the middle of March, human activity slowed down-- >> Yeah. >> In a way that no one could have forecast ever on a single event, globally which is just fascinating. And you think of the amount of airplanes that were not flying and trains that we're not moving and people not moving. Did you have any any data or have you been able to collect data or see data as the impact of that? Not only directly in wherever the sensors are, but a kind of a second order impact because of the lack of pollution and the other kind of human activity that just went down. I mean, certainly a lot of memes (Bryan chuckling) on social media of all the animals-- >> Yeah. >> Come back into the city. But I'm just curious if you have any data in the observation? >> Yeah, we're part of actually a global study, there's couple of hundred of us that are contributing our data to what we call the Silent Cities project. It's being coordinated out of Europe right now. So we placed our sensors out in different areas, actually around West Lafayette area here in Indiana, near road crossings and that sort of thing to be able to kind of capture that information. We have had in this area here now, the 17 year study. So we do have studies that get into areas that tend to be fairly urban. So we do have a lot of information. I tell you, I don't need my sensors to tell me something that I already know and you suspect is true. Our cities were quiet, much quieter during the COVID situation. And it's continued to kind of get a little bit louder, as we've kind of released some of the policies that put us into our homes. And so yes, there is a major change. Now there have been a couple of studies that just come out that are pretty interesting. One, which was in San Francisco looking at the white-crowned sparrow. And they looked at historical data that went back something like 20 years. And they found that the birds in the cities were singing a much softer, 30% softer. >> Really? >> And they, yeah, and they would lower their frequencies. So the way sound works is that if you lower your frequencies that sound can travel farther. And so the males can now hear themselves twice as far just due to the fact that our cities are quieter. So it does have an impact on animals, truly it does. There was some studies back in 2001, during  the September, the 9/11 crisis as well, where people are going out and kind of looking at data, acoustic data, and discovering that things were much quieter. I'd be very interested to look at some of the data we have in our oceans, to what extent are oceans quieter. Our oceans sadly are the loudest part of this planet. It's really noisy, sound travels, five times farther. Generally the noise is lower frequencies, and we have lots of ships that are all over the planet and in our oceans. So I'd really be interested in those kinds of studies as well, to what extent is it impacting and helping our friends in the oceans. >> Right, right, well, I was just going to ask you that question because I think a lot of people clearly understand sound in the air that surrounds us, but you talk a lot about sound in ocean, and sound as an indicator of ocean health, and again, this concept of a chorus. And I think everybody's probably familiar with the sounds of the humpback whale right? He got very popular and we've all seen and heard that. But you're doing a lot of research, as you said, in oceans and in water. And I wonder if you can, again, kind of provide a little bit more color around that, because I don't think you people, maybe we're just not that tuned into it, think of the ocean or water as a rich sound environment especially to the degree as you're talking about where you can actually start to really understand what's going on. >> Yeah, I mean, some of us think that sound in the oceans is probably more important to animals than on land, on the terrestrial side. Sound helps animals to navigate through complex waterways and find food resources. You can only use site so far underwater especially when it gets to be kind of dark, once you get down to certain levels. So there many of us think that sound is probably going to be an important component to measuring the status of health in our oceans. >> It's great. Well, Bryan, I really enjoyed this conversation. I've really enjoyed your Ted Talk, and now I've got a bunch of research papers I want to dig into a little bit more as well. >> Okay.(chuckling) >> It's a fascinating topic, but I think the most important thing that you talked about extensively in your Ted Talk is really just taking a minute to take a step back from the individual perspective, appreciate what's around us, hear, that information and I think there's a real direct correlation to the power of exascale, to the power of hearing this data, processing this data, and putting intelligence on that data, understanding that data in a good way, in a positive way, in a delightful way, spiritual way, even that we couldn't do before, or we just weren't paying attention like with what you know is on your phone please-- >> Yeah, really. >> It's all around you. It's been there a whole time. >> Yeah. (both chuckling) >> Yeah, Jeff, I really encourage your viewers to count it, just go out and listen. As we say, go out and listen and join the mission. >> I love it, and you can get started by going to the Center for Global Soundscapes and you have a beautiful landscape. I had it going earlier this morning while I was digging through some of the research of Bryan. (Bryan chuckling) Thank you very much (Bryan murmurs) and really enjoyed the conversation best to you-- >> Okay. >> And your team and your continued success. >> Alright, thank you. >> Alright, thank you. All right, he's Bryan-- >> Goodbye. >> I'm Jeff, you're watching theCUBE. (Bryan chuckling) for continuing coverage of Exascale Day. Thanks for watching. We'll see you next time. (calm ambient music)

>>don't talk mhm, >>Okay, thing is my presentation on coherent nonlinear dynamics and combinatorial optimization. This is going to be a talk to introduce an approach we're taking to the analysis of the performance of coherent using machines. So let me start with a brief introduction to easing optimization. The easing model represents a set of interacting magnetic moments or spins the total energy given by the expression shown at the bottom left of this slide. Here, the signal variables are meditate binary values. The Matrix element J. I. J. Represents the interaction, strength and signed between any pair of spins. I. J and A Chive represents a possible local magnetic field acting on each thing. The easing ground state problem is to find an assignment of binary spin values that achieves the lowest possible value of total energy. And an instance of the easing problem is specified by giving numerical values for the Matrix J in Vector H. Although the easy model originates in physics, we understand the ground state problem to correspond to what would be called quadratic binary optimization in the field of operations research and in fact, in terms of computational complexity theory, it could be established that the easing ground state problem is np complete. Qualitatively speaking, this makes the easing problem a representative sort of hard optimization problem, for which it is expected that the runtime required by any computational algorithm to find exact solutions should, as anatomically scale exponentially with the number of spends and for worst case instances at each end. Of course, there's no reason to believe that the problem instances that actually arrives in practical optimization scenarios are going to be worst case instances. And it's also not generally the case in practical optimization scenarios that we demand absolute optimum solutions. Usually we're more interested in just getting the best solution we can within an affordable cost, where costs may be measured in terms of time, service fees and or energy required for a computation. This focuses great interest on so called heuristic algorithms for the easing problem in other NP complete problems which generally get very good but not guaranteed optimum solutions and run much faster than algorithms that are designed to find absolute Optima. To get some feeling for present day numbers, we can consider the famous traveling salesman problem for which extensive compilations of benchmarking data may be found online. A recent study found that the best known TSP solver required median run times across the Library of Problem instances That scaled is a very steep route exponential for end up to approximately 4500. This gives some indication of the change in runtime scaling for generic as opposed the worst case problem instances. Some of the instances considered in this study were taken from a public library of T SPS derived from real world Veil aside design data. This feels I TSP Library includes instances within ranging from 131 to 744,710 instances from this library with end between 6880 13,584 were first solved just a few years ago in 2017 requiring days of run time and a 48 core to King hurts cluster, while instances with and greater than or equal to 14,233 remain unsolved exactly by any means. Approximate solutions, however, have been found by heuristic methods for all instances in the VLS i TSP library with, for example, a solution within 0.14% of a no lower bound, having been discovered, for instance, with an equal 19,289 requiring approximately two days of run time on a single core of 2.4 gigahertz. Now, if we simple mindedly extrapolate the root exponential scaling from the study up to an equal 4500, we might expect that an exact solver would require something more like a year of run time on the 48 core cluster used for the N equals 13,580 for instance, which shows how much a very small concession on the quality of the solution makes it possible to tackle much larger instances with much lower cost. At the extreme end, the largest TSP ever solved exactly has an equal 85,900. This is an instance derived from 19 eighties VLSI design, and it's required 136 CPU. Years of computation normalized to a single cord, 2.4 gigahertz. But the 24 larger so called world TSP benchmark instance within equals 1,904,711 has been solved approximately within ophthalmology. Gap bounded below 0.474%. Coming back to the general. Practical concerns have applied optimization. We may note that a recent meta study analyzed the performance of no fewer than 37 heuristic algorithms for Max cut and quadratic pioneer optimization problems and found the performance sort and found that different heuristics work best for different problem instances selected from a large scale heterogeneous test bed with some evidence but cryptic structure in terms of what types of problem instances were best solved by any given heuristic. Indeed, their their reasons to believe that these results from Mexico and quadratic binary optimization reflected general principle of performance complementarity among heuristic optimization algorithms in the practice of solving heart optimization problems there. The cerise is a critical pre processing issue of trying to guess which of a number of available good heuristic algorithms should be chosen to tackle a given problem. Instance, assuming that any one of them would incur high costs to run on a large problem, instances incidence, making an astute choice of heuristic is a crucial part of maximizing overall performance. Unfortunately, we still have very little conceptual insight about what makes a specific problem instance, good or bad for any given heuristic optimization algorithm. This has certainly been pinpointed by researchers in the field is a circumstance that must be addressed. So adding this all up, we see that a critical frontier for cutting edge academic research involves both the development of novel heuristic algorithms that deliver better performance, with lower cost on classes of problem instances that are underserved by existing approaches, as well as fundamental research to provide deep conceptual insight into what makes a given problem in, since easy or hard for such algorithms. In fact, these days, as we talk about the end of Moore's law and speculate about a so called second quantum revolution, it's natural to talk not only about novel algorithms for conventional CPUs but also about highly customized special purpose hardware architectures on which we may run entirely unconventional algorithms for combinatorial optimization such as easing problem. So against that backdrop, I'd like to use my remaining time to introduce our work on analysis of coherent using machine architectures and associate ID optimization algorithms. These machines, in general, are a novel class of information processing architectures for solving combinatorial optimization problems by embedding them in the dynamics of analog, physical or cyber physical systems, in contrast to both MAWR traditional engineering approaches that build using machines using conventional electron ICS and more radical proposals that would require large scale quantum entanglement. The emerging paradigm of coherent easing machines leverages coherent nonlinear dynamics in photonic or Opto electronic platforms to enable near term construction of large scale prototypes that leverage post Simoes information dynamics, the general structure of of current CM systems has shown in the figure on the right. The role of the easing spins is played by a train of optical pulses circulating around a fiber optical storage ring. A beam splitter inserted in the ring is used to periodically sample the amplitude of every optical pulse, and the measurement results are continually read into a refugee A, which uses them to compute perturbations to be applied to each pulse by a synchronized optical injections. These perturbations, air engineered to implement the spin, spin coupling and local magnetic field terms of the easing Hamiltonian, corresponding to a linear part of the CME Dynamics, a synchronously pumped parametric amplifier denoted here as PPL and Wave Guide adds a crucial nonlinear component to the CIA and Dynamics as well. In the basic CM algorithm, the pump power starts very low and has gradually increased at low pump powers. The amplitude of the easing spin pulses behaviors continuous, complex variables. Who Israel parts which can be positive or negative, play the role of play the role of soft or perhaps mean field spins once the pump, our crosses the threshold for parametric self oscillation. In the optical fiber ring, however, the attitudes of the easing spin pulses become effectively Qantas ized into binary values while the pump power is being ramped up. The F P J subsystem continuously applies its measurement based feedback. Implementation of the using Hamiltonian terms, the interplay of the linear rised using dynamics implemented by the F P G A and the threshold conversation dynamics provided by the sink pumped Parametric amplifier result in the final state of the optical optical pulse amplitude at the end of the pump ramp that could be read as a binary strain, giving a proposed solution of the easing ground state problem. This method of solving easing problem seems quite different from a conventional algorithm that runs entirely on a digital computer as a crucial aspect of the computation is performed physically by the analog, continuous, coherent, nonlinear dynamics of the optical degrees of freedom. In our efforts to analyze CIA and performance, we have therefore turned to the tools of dynamical systems theory, namely, a study of modifications, the evolution of critical points and apologies of hetero clinic orbits and basins of attraction. We conjecture that such analysis can provide fundamental insight into what makes certain optimization instances hard or easy for coherent using machines and hope that our approach can lead to both improvements of the course, the AM algorithm and a pre processing rubric for rapidly assessing the CME suitability of new instances. Okay, to provide a bit of intuition about how this all works, it may help to consider the threshold dynamics of just one or two optical parametric oscillators in the CME architecture just described. We can think of each of the pulse time slots circulating around the fiber ring, as are presenting an independent Opio. We can think of a single Opio degree of freedom as a single, resonant optical node that experiences linear dissipation, do toe out coupling loss and gain in a pump. Nonlinear crystal has shown in the diagram on the upper left of this slide as the pump power is increased from zero. As in the CME algorithm, the non linear game is initially to low toe overcome linear dissipation, and the Opio field remains in a near vacuum state at a critical threshold. Value gain. Equal participation in the Popeo undergoes a sort of lazing transition, and the study states of the OPIO above this threshold are essentially coherent states. There are actually two possible values of the Opio career in amplitude and any given above threshold pump power which are equal in magnitude but opposite in phase when the OPI across the special diet basically chooses one of the two possible phases randomly, resulting in the generation of a single bit of information. If we consider to uncoupled, Opio has shown in the upper right diagram pumped it exactly the same power at all times. Then, as the pump power has increased through threshold, each Opio will independently choose the phase and thus to random bits are generated for any number of uncoupled. Oppose the threshold power per opio is unchanged from the single Opio case. Now, however, consider a scenario in which the two appeals air, coupled to each other by a mutual injection of their out coupled fields has shown in the diagram on the lower right. One can imagine that depending on the sign of the coupling parameter Alfa, when one Opio is lazing, it will inject a perturbation into the other that may interfere either constructively or destructively, with the feel that it is trying to generate by its own lazing process. As a result, when came easily showed that for Alfa positive, there's an effective ferro magnetic coupling between the two Opio fields and their collective oscillation threshold is lowered from that of the independent Opio case. But on Lee for the two collective oscillation modes in which the two Opio phases are the same for Alfa Negative, the collective oscillation threshold is lowered on Lee for the configurations in which the Opio phases air opposite. So then, looking at how Alfa is related to the J. I. J matrix of the easing spin coupling Hamiltonian, it follows that we could use this simplistic to a p o. C. I am to solve the ground state problem of a fair magnetic or anti ferro magnetic ankles to easing model simply by increasing the pump power from zero and observing what phase relation occurs as the two appeals first start delays. Clearly, we can imagine generalizing this story toe larger, and however the story doesn't stay is clean and simple for all larger problem instances. And to find a more complicated example, we only need to go to n equals four for some choices of J J for n equals, for the story remains simple. Like the n equals two case. The figure on the upper left of this slide shows the energy of various critical points for a non frustrated and equals, for instance, in which the first bifurcated critical point that is the one that I forget to the lowest pump value a. Uh, this first bifurcated critical point flows as symptomatically into the lowest energy easing solution and the figure on the upper right. However, the first bifurcated critical point flows to a very good but sub optimal minimum at large pump power. The global minimum is actually given by a distinct critical critical point that first appears at a higher pump power and is not automatically connected to the origin. The basic C am algorithm is thus not able to find this global minimum. Such non ideal behaviors needs to become more confident. Larger end for the n equals 20 instance, showing the lower plots where the lower right plot is just a zoom into a region of the lower left lot. It can be seen that the global minimum corresponds to a critical point that first appears out of pump parameter, a around 0.16 at some distance from the idiomatic trajectory of the origin. That's curious to note that in both of these small and examples, however, the critical point corresponding to the global minimum appears relatively close to the idiomatic projector of the origin as compared to the most of the other local minima that appear. We're currently working to characterize the face portrait topology between the global minimum in the antibiotic trajectory of the origin, taking clues as to how the basic C am algorithm could be generalized to search for non idiomatic trajectories that jump to the global minimum during the pump ramp. Of course, n equals 20 is still too small to be of interest for practical optimization applications. But the advantage of beginning with the study of small instances is that we're able reliably to determine their global minima and to see how they relate to the 80 about trajectory of the origin in the basic C am algorithm. In the smaller and limit, we can also analyze fully quantum mechanical models of Syrian dynamics. But that's a topic for future talks. Um, existing large scale prototypes are pushing into the range of in equals 10 to the 4 10 to 5 to six. So our ultimate objective in theoretical analysis really has to be to try to say something about CIA and dynamics and regime of much larger in our initial approach to characterizing CIA and behavior in the large in regime relies on the use of random matrix theory, and this connects to prior research on spin classes, SK models and the tap equations etcetera. At present, we're focusing on statistical characterization of the CIA ingredient descent landscape, including the evolution of critical points in their Eigen value spectra. As the pump power is gradually increased. We're investigating, for example, whether there could be some way to exploit differences in the relative stability of the global minimum versus other local minima. We're also working to understand the deleterious or potentially beneficial effects of non ideologies, such as a symmetry in the implemented these and couplings. Looking one step ahead, we plan to move next in the direction of considering more realistic classes of problem instances such as quadratic, binary optimization with constraints. Eso In closing, I should acknowledge people who did the hard work on these things that I've shown eso. My group, including graduate students Ed winning, Daniel Wennberg, Tatsuya Nagamoto and Atsushi Yamamura, have been working in close collaboration with Syria Ganguly, Marty Fair and Amir Safarini Nini, all of us within the Department of Applied Physics at Stanford University. On also in collaboration with the Oshima Moto over at NTT 55 research labs, Onda should acknowledge funding support from the NSF by the Coherent Easing Machines Expedition in computing, also from NTT five research labs, Army Research Office and Exxon Mobil. Uh, that's it. Thanks very much. >>Mhm e >>t research and the Oshie for putting together this program and also the opportunity to speak here. My name is Al Gore ism or Andy and I'm from Caltech, and today I'm going to tell you about the work that we have been doing on networks off optical parametric oscillators and how we have been using them for icing machines and how we're pushing them toward Cornum photonics to acknowledge my team at Caltech, which is now eight graduate students and five researcher and postdocs as well as collaborators from all over the world, including entity research and also the funding from different places, including entity. So this talk is primarily about networks of resonate er's, and these networks are everywhere from nature. For instance, the brain, which is a network of oscillators all the way to optics and photonics and some of the biggest examples or metal materials, which is an array of small resonate er's. And we're recently the field of technological photonics, which is trying thio implement a lot of the technological behaviors of models in the condensed matter, physics in photonics and if you want to extend it even further, some of the implementations off quantum computing are technically networks of quantum oscillators. So we started thinking about these things in the context of icing machines, which is based on the icing problem, which is based on the icing model, which is the simple summation over the spins and spins can be their upward down and the couplings is given by the JJ. And the icing problem is, if you know J I J. What is the spin configuration that gives you the ground state? And this problem is shown to be an MP high problem. So it's computational e important because it's a representative of the MP problems on NPR. Problems are important because first, their heart and standard computers if you use a brute force algorithm and they're everywhere on the application side. That's why there is this demand for making a machine that can target these problems, and hopefully it can provide some meaningful computational benefit compared to the standard digital computers. So I've been building these icing machines based on this building block, which is a degenerate optical parametric. Oscillator on what it is is resonator with non linearity in it, and we pump these resonate er's and we generate the signal at half the frequency of the pump. One vote on a pump splits into two identical photons of signal, and they have some very interesting phase of frequency locking behaviors. And if you look at the phase locking behavior, you realize that you can actually have two possible phase states as the escalation result of these Opio which are off by pie, and that's one of the important characteristics of them. So I want to emphasize a little more on that and I have this mechanical analogy which are basically two simple pendulum. But there are parametric oscillators because I'm going to modulate the parameter of them in this video, which is the length of the string on by that modulation, which is that will make a pump. I'm gonna make a muscular. That'll make a signal which is half the frequency of the pump. And I have two of them to show you that they can acquire these face states so they're still facing frequency lock to the pump. But it can also lead in either the zero pie face states on. The idea is to use this binary phase to represent the binary icing spin. So each opio is going to represent spin, which can be either is your pie or up or down. And to implement the network of these resonate er's, we use the time off blood scheme, and the idea is that we put impulses in the cavity. These pulses air separated by the repetition period that you put in or t r. And you can think about these pulses in one resonator, xaz and temporarily separated synthetic resonate Er's if you want a couple of these resonator is to each other, and now you can introduce these delays, each of which is a multiple of TR. If you look at the shortest delay it couples resonator wanted to 2 to 3 and so on. If you look at the second delay, which is two times a rotation period, the couple's 123 and so on. And if you have and minus one delay lines, then you can have any potential couplings among these synthetic resonate er's. And if I can introduce these modulators in those delay lines so that I can strength, I can control the strength and the phase of these couplings at the right time. Then I can have a program will all toe all connected network in this time off like scheme, and the whole physical size of the system scales linearly with the number of pulses. So the idea of opium based icing machine is didn't having these o pos, each of them can be either zero pie and I can arbitrarily connect them to each other. And then I start with programming this machine to a given icing problem by just setting the couplings and setting the controllers in each of those delight lines. So now I have a network which represents an icing problem. Then the icing problem maps to finding the face state that satisfy maximum number of coupling constraints. And the way it happens is that the icing Hamiltonian maps to the linear loss of the network. And if I start adding gain by just putting pump into the network, then the OPI ohs are expected to oscillate in the lowest, lowest lost state. And, uh and we have been doing these in the past, uh, six or seven years and I'm just going to quickly show you the transition, especially what happened in the first implementation, which was using a free space optical system and then the guided wave implementation in 2016 and the measurement feedback idea which led to increasing the size and doing actual computation with these machines. So I just want to make this distinction here that, um, the first implementation was an all optical interaction. We also had an unequal 16 implementation. And then we transition to this measurement feedback idea, which I'll tell you quickly what it iss on. There's still a lot of ongoing work, especially on the entity side, to make larger machines using the measurement feedback. But I'm gonna mostly focused on the all optical networks and how we're using all optical networks to go beyond simulation of icing Hamiltonian both in the linear and non linear side and also how we're working on miniaturization of these Opio networks. So the first experiment, which was the four opium machine, it was a free space implementation and this is the actual picture off the machine and we implemented a small and it calls for Mexico problem on the machine. So one problem for one experiment and we ran the machine 1000 times, we looked at the state and we always saw it oscillate in one of these, um, ground states of the icing laboratoria. So then the measurement feedback idea was to replace those couplings and the controller with the simulator. So we basically simulated all those coherent interactions on on FB g. A. And we replicated the coherent pulse with respect to all those measurements. And then we injected it back into the cavity and on the near to you still remain. So it still is a non. They're dynamical system, but the linear side is all simulated. So there are lots of questions about if this system is preserving important information or not, or if it's gonna behave better. Computational wars. And that's still ah, lot of ongoing studies. But nevertheless, the reason that this implementation was very interesting is that you don't need the end minus one delight lines so you can just use one. Then you can implement a large machine, and then you can run several thousands of problems in the machine, and then you can compare the performance from the computational perspective Looks so I'm gonna split this idea of opium based icing machine into two parts. One is the linear part, which is if you take out the non linearity out of the resonator and just think about the connections. You can think about this as a simple matrix multiplication scheme. And that's basically what gives you the icing Hambletonian modeling. So the optical laws of this network corresponds to the icing Hamiltonian. And if I just want to show you the example of the n equals for experiment on all those face states and the history Graham that we saw, you can actually calculate the laws of each of those states because all those interferences in the beam splitters and the delay lines are going to give you a different losses. And then you will see that the ground states corresponds to the lowest laws of the actual optical network. If you add the non linearity, the simple way of thinking about what the non linearity does is that it provides to gain, and then you start bringing up the gain so that it hits the loss. Then you go through the game saturation or the threshold which is going to give you this phase bifurcation. So you go either to zero the pie face state. And the expectation is that Theis, the network oscillates in the lowest possible state, the lowest possible loss state. There are some challenges associated with this intensity Durban face transition, which I'm going to briefly talk about. I'm also going to tell you about other types of non aerodynamics that we're looking at on the non air side of these networks. So if you just think about the linear network, we're actually interested in looking at some technological behaviors in these networks. And the difference between looking at the technological behaviors and the icing uh, machine is that now, First of all, we're looking at the type of Hamilton Ian's that are a little different than the icing Hamilton. And one of the biggest difference is is that most of these technological Hamilton Ian's that require breaking the time reversal symmetry, meaning that you go from one spin to in the one side to another side and you get one phase. And if you go back where you get a different phase, and the other thing is that we're not just interested in finding the ground state, we're actually now interesting and looking at all sorts of states and looking at the dynamics and the behaviors of all these states in the network. So we started with the simplest implementation, of course, which is a one d chain of thes resonate, er's, which corresponds to a so called ssh model. In the technological work, we get the similar energy to los mapping and now we can actually look at the band structure on. This is an actual measurement that we get with this associate model and you see how it reasonably how How? Well, it actually follows the prediction and the theory. One of the interesting things about the time multiplexing implementation is that now you have the flexibility of changing the network as you are running the machine. And that's something unique about this time multiplex implementation so that we can actually look at the dynamics. And one example that we have looked at is we can actually go through the transition off going from top A logical to the to the standard nontrivial. I'm sorry to the trivial behavior of the network. You can then look at the edge states and you can also see the trivial and states and the technological at states actually showing up in this network. We have just recently implement on a two D, uh, network with Harper Hofstadter model and when you don't have the results here. But we're one of the other important characteristic of time multiplexing is that you can go to higher and higher dimensions and keeping that flexibility and dynamics, and we can also think about adding non linearity both in a classical and quantum regimes, which is going to give us a lot of exotic, no classical and quantum, non innate behaviors in these networks. Yeah, So I told you about the linear side. Mostly let me just switch gears and talk about the nonlinear side of the network. And the biggest thing that I talked about so far in the icing machine is this face transition that threshold. So the low threshold we have squeezed state in these. Oh, pios, if you increase the pump, we go through this intensity driven phase transition and then we got the face stays above threshold. And this is basically the mechanism off the computation in these O pos, which is through this phase transition below to above threshold. So one of the characteristics of this phase transition is that below threshold, you expect to see quantum states above threshold. You expect to see more classical states or coherent states, and that's basically corresponding to the intensity off the driving pump. So it's really hard to imagine that it can go above threshold. Or you can have this friends transition happen in the all in the quantum regime. And there are also some challenges associated with the intensity homogeneity off the network, which, for example, is if one opioid starts oscillating and then its intensity goes really high. Then it's going to ruin this collective decision making off the network because of the intensity driven face transition nature. So So the question is, can we look at other phase transitions? Can we utilize them for both computing? And also can we bring them to the quantum regime on? I'm going to specifically talk about the face transition in the spectral domain, which is the transition from the so called degenerate regime, which is what I mostly talked about to the non degenerate regime, which happens by just tuning the phase of the cavity. And what is interesting is that this phase transition corresponds to a distinct phase noise behavior. So in the degenerate regime, which we call it the order state, you're gonna have the phase being locked to the phase of the pump. As I talked about non degenerate regime. However, the phase is the phase is mostly dominated by the quantum diffusion. Off the off the phase, which is limited by the so called shallow towns limit, and you can see that transition from the general to non degenerate, which also has distinct symmetry differences. And this transition corresponds to a symmetry breaking in the non degenerate case. The signal can acquire any of those phases on the circle, so it has a you one symmetry. Okay, and if you go to the degenerate case, then that symmetry is broken and you only have zero pie face days I will look at. So now the question is can utilize this phase transition, which is a face driven phase transition, and can we use it for similar computational scheme? So that's one of the questions that were also thinking about. And it's not just this face transition is not just important for computing. It's also interesting from the sensing potentials and this face transition, you can easily bring it below threshold and just operated in the quantum regime. Either Gaussian or non Gaussian. If you make a network of Opio is now, we can see all sorts off more complicated and more interesting phase transitions in the spectral domain. One of them is the first order phase transition, which you get by just coupling to Opio, and that's a very abrupt face transition and compared to the to the single Opio phase transition. And if you do the couplings right, you can actually get a lot of non her mission dynamics and exceptional points, which are actually very interesting to explore both in the classical and quantum regime. And I should also mention that you can think about the cup links to be also nonlinear couplings. And that's another behavior that you can see, especially in the nonlinear in the non degenerate regime. So with that, I basically told you about these Opio networks, how we can think about the linear scheme and the linear behaviors and how we can think about the rich, nonlinear dynamics and non linear behaviors both in the classical and quantum regime. I want to switch gear and tell you a little bit about the miniaturization of these Opio networks. And of course, the motivation is if you look at the electron ICS and what we had 60 or 70 years ago with vacuum tube and how we transition from relatively small scale computers in the order of thousands of nonlinear elements to billions of non elements where we are now with the optics is probably very similar to 70 years ago, which is a table talk implementation. And the question is, how can we utilize nano photonics? I'm gonna just briefly show you the two directions on that which we're working on. One is based on lithium Diabate, and the other is based on even a smaller resonate er's could you? So the work on Nana Photonic lithium naive. It was started in collaboration with Harvard Marko Loncar, and also might affair at Stanford. And, uh, we could show that you can do the periodic polling in the phenomenon of it and get all sorts of very highly nonlinear processes happening in this net. Photonic periodically polls if, um Diabate. And now we're working on building. Opio was based on that kind of photonic the film Diabate. And these air some some examples of the devices that we have been building in the past few months, which I'm not gonna tell you more about. But the O. P. O. S. And the Opio Networks are in the works. And that's not the only way of making large networks. Um, but also I want to point out that The reason that these Nana photonic goblins are actually exciting is not just because you can make a large networks and it can make him compact in a in a small footprint. They also provide some opportunities in terms of the operation regime. On one of them is about making cat states and Opio, which is, can we have the quantum superposition of the zero pie states that I talked about and the Net a photonic within? I've It provides some opportunities to actually get closer to that regime because of the spatial temporal confinement that you can get in these wave guides. So we're doing some theory on that. We're confident that the type of non linearity two losses that it can get with these platforms are actually much higher than what you can get with other platform their existing platforms and to go even smaller. We have been asking the question off. What is the smallest possible Opio that you can make? Then you can think about really wavelength scale type, resonate er's and adding the chi to non linearity and see how and when you can get the Opio to operate. And recently, in collaboration with us see, we have been actually USC and Creole. We have demonstrated that you can use nano lasers and get some spin Hamilton and implementations on those networks. So if you can build the a P. O s, we know that there is a path for implementing Opio Networks on on such a nano scale. So we have looked at these calculations and we try to estimate the threshold of a pos. Let's say for me resonator and it turns out that it can actually be even lower than the type of bulk Pip Llano Pos that we have been building in the past 50 years or so. So we're working on the experiments and we're hoping that we can actually make even larger and larger scale Opio networks. So let me summarize the talk I told you about the opium networks and our work that has been going on on icing machines and the measurement feedback. And I told you about the ongoing work on the all optical implementations both on the linear side and also on the nonlinear behaviors. And I also told you a little bit about the efforts on miniaturization and going to the to the Nano scale. So with that, I would like Thio >>three from the University of Tokyo. Before I thought that would like to thank you showing all the stuff of entity for the invitation and the organization of this online meeting and also would like to say that it has been very exciting to see the growth of this new film lab. And I'm happy to share with you today of some of the recent works that have been done either by me or by character of Hong Kong. Honest Group indicates the title of my talk is a neuro more fic in silica simulator for the communities in machine. And here is the outline I would like to make the case that the simulation in digital Tektronix of the CME can be useful for the better understanding or improving its function principles by new job introducing some ideas from neural networks. This is what I will discuss in the first part and then it will show some proof of concept of the game and performance that can be obtained using dissimulation in the second part and the protection of the performance that can be achieved using a very large chaos simulator in the third part and finally talk about future plans. So first, let me start by comparing recently proposed izing machines using this table there is elected from recent natural tronics paper from the village Park hard people, and this comparison shows that there's always a trade off between energy efficiency, speed and scalability that depends on the physical implementation. So in red, here are the limitation of each of the servers hardware on, interestingly, the F p G, a based systems such as a producer, digital, another uh Toshiba beautification machine or a recently proposed restricted Bozeman machine, FPD A by a group in Berkeley. They offer a good compromise between speed and scalability. And this is why, despite the unique advantage that some of these older hardware have trust as the currency proposition in Fox, CBS or the energy efficiency off memory Sisters uh P. J. O are still an attractive platform for building large organizing machines in the near future. The reason for the good performance of Refugee A is not so much that they operate at the high frequency. No, there are particular in use, efficient, but rather that the physical wiring off its elements can be reconfigured in a way that limits the funding human bottleneck, larger, funny and phenols and the long propagation video information within the system. In this respect, the LPGA is They are interesting from the perspective off the physics off complex systems, but then the physics of the actions on the photos. So to put the performance of these various hardware and perspective, we can look at the competition of bringing the brain the brain complete, using billions of neurons using only 20 watts of power and operates. It's a very theoretically slow, if we can see and so this impressive characteristic, they motivate us to try to investigate. What kind of new inspired principles be useful for designing better izing machines? The idea of this research project in the future collaboration it's to temporary alleviates the limitations that are intrinsic to the realization of an optical cortex in machine shown in the top panel here. By designing a large care simulator in silicone in the bottom here that can be used for digesting the better organization principles of the CIA and this talk, I will talk about three neuro inspired principles that are the symmetry of connections, neural dynamics orphan chaotic because of symmetry, is interconnectivity the infrastructure? No. Next talks are not composed of the reputation of always the same types of non environments of the neurons, but there is a local structure that is repeated. So here's the schematic of the micro column in the cortex. And lastly, the Iraqi co organization of connectivity connectivity is organizing a tree structure in the brain. So here you see a representation of the Iraqi and organization of the monkey cerebral cortex. So how can these principles we used to improve the performance of the icing machines? And it's in sequence stimulation. So, first about the two of principles of the estimate Trian Rico structure. We know that the classical approximation of the car testing machine, which is the ground toe, the rate based on your networks. So in the case of the icing machines, uh, the okay, Scott approximation can be obtained using the trump active in your position, for example, so the times of both of the system they are, they can be described by the following ordinary differential equations on in which, in case of see, I am the X, I represent the in phase component of one GOP Oh, Theo f represents the monitor optical parts, the district optical Parametric amplification and some of the good I JoJo extra represent the coupling, which is done in the case of the measure of feedback coupling cm using oh, more than detection and refugee A and then injection off the cooking time and eso this dynamics in both cases of CNN in your networks, they can be written as the grand set of a potential function V, and this written here, and this potential functionally includes the rising Maccagnan. So this is why it's natural to use this type of, uh, dynamics to solve the icing problem in which the Omega I J or the eyes in coping and the H is the extension of the icing and attorney in India and expect so. Not that this potential function can only be defined if the Omega I j. R. A. Symmetric. So the well known problem of this approach is that this potential function V that we obtain is very non convicts at low temperature, and also one strategy is to gradually deformed this landscape, using so many in process. But there is no theorem. Unfortunately, that granted conventions to the global minimum of There's even Tony and using this approach. And so this is why we propose, uh, to introduce a macro structures of the system where one analog spin or one D O. P. O is replaced by a pair off one another spin and one error, according viable. And the addition of this chemical structure introduces a symmetry in the system, which in terms induces chaotic dynamics, a chaotic search rather than a learning process for searching for the ground state of the icing. Every 20 within this massacre structure the role of the er variable eyes to control the amplitude off the analog spins toe force. The amplitude of the expense toe become equal to certain target amplitude a uh and, uh, and this is done by modulating the strength off the icing complaints or see the the error variable E I multiply the icing complaint here in the dynamics off air d o p. O. On then the dynamics. The whole dynamics described by this coupled equations because the e I do not necessarily take away the same value for the different. I thesis introduces a symmetry in the system, which in turn creates security dynamics, which I'm sure here for solving certain current size off, um, escape problem, Uh, in which the X I are shown here and the i r from here and the value of the icing energy showing the bottom plots. You see this Celtics search that visit various local minima of the as Newtonian and eventually finds the global minimum? Um, it can be shown that this modulation off the target opportunity can be used to destabilize all the local minima off the icing evertonians so that we're gonna do not get stuck in any of them. On more over the other types of attractors I can eventually appear, such as limits I contractors, Okot contractors. They can also be destabilized using the motivation of the target and Batuta. And so we have proposed in the past two different moderation of the target amateur. The first one is a modulation that ensure the uh 100 reproduction rate of the system to become positive on this forbids the creation off any nontrivial tractors. And but in this work, I will talk about another moderation or arrested moderation which is given here. That works, uh, as well as this first uh, moderation, but is easy to be implemented on refugee. So this couple of the question that represent becoming the stimulation of the cortex in machine with some error correction they can be implemented especially efficiently on an F B. G. And here I show the time that it takes to simulate three system and also in red. You see, at the time that it takes to simulate the X I term the EI term, the dot product and the rising Hamiltonian for a system with 500 spins and Iraq Spain's equivalent to 500 g. O. P. S. So >>in >>f b d a. The nonlinear dynamics which, according to the digital optical Parametric amplification that the Opa off the CME can be computed in only 13 clock cycles at 300 yards. So which corresponds to about 0.1 microseconds. And this is Toby, uh, compared to what can be achieved in the measurements back O C. M. In which, if we want to get 500 timer chip Xia Pios with the one she got repetition rate through the obstacle nine narrative. Uh, then way would require 0.5 microseconds toe do this so the submission in F B J can be at least as fast as ah one g repression. Uh, replicate pulsed laser CIA Um, then the DOT product that appears in this differential equation can be completed in 43 clock cycles. That's to say, one microseconds at 15 years. So I pieced for pouring sizes that are larger than 500 speeds. The dot product becomes clearly the bottleneck, and this can be seen by looking at the the skating off the time the numbers of clock cycles a text to compute either the non in your optical parts or the dog products, respect to the problem size. And And if we had infinite amount of resources and PGA to simulate the dynamics, then the non illogical post can could be done in the old one. On the mattress Vector product could be done in the low carrot off, located off scales as a look at it off and and while the guide off end. Because computing the dot product involves assuming all the terms in the product, which is done by a nephew, GE by another tree, which heights scarce logarithmic any with the size of the system. But This is in the case if we had an infinite amount of resources on the LPGA food, but for dealing for larger problems off more than 100 spins. Usually we need to decompose the metrics into ah, smaller blocks with the block side that are not you here. And then the scaling becomes funny, non inner parts linear in the end, over you and for the products in the end of EU square eso typically for low NF pdf cheap PGA you the block size off this matrix is typically about 100. So clearly way want to make you as large as possible in order to maintain this scanning in a log event for the numbers of clock cycles needed to compute the product rather than this and square that occurs if we decompose the metrics into smaller blocks. But the difficulty in, uh, having this larger blocks eyes that having another tree very large Haider tree introduces a large finding and finance and long distance start a path within the refugee. So the solution to get higher performance for a simulator of the contest in machine eyes to get rid of this bottleneck for the dot product by increasing the size of this at the tree. And this can be done by organizing your critique the electrical components within the LPGA in order which is shown here in this, uh, right panel here in order to minimize the finding finance of the system and to minimize the long distance that a path in the in the fpt So I'm not going to the details of how this is implemented LPGA. But just to give you a idea off why the Iraqi Yahiko organization off the system becomes the extremely important toe get good performance for similar organizing machine. So instead of instead of getting into the details of the mpg implementation, I would like to give some few benchmark results off this simulator, uh, off the that that was used as a proof of concept for this idea which is can be found in this archive paper here and here. I should results for solving escape problems. Free connected person, randomly person minus one spring last problems and we sure, as we use as a metric the numbers of the mattress Victor products since it's the bottleneck of the computation, uh, to get the optimal solution of this escape problem with the Nina successful BT against the problem size here and and in red here, this propose FDJ implementation and in ah blue is the numbers of retrospective product that are necessary for the C. I am without error correction to solve this escape programs and in green here for noisy means in an evening which is, uh, behavior with similar to the Cartesian mission. Uh, and so clearly you see that the scaring off the numbers of matrix vector product necessary to solve this problem scales with a better exponents than this other approaches. So So So that's interesting feature of the system and next we can see what is the real time to solution to solve this SK instances eso in the last six years, the time institution in seconds to find a grand state of risk. Instances remain answers probability for different state of the art hardware. So in red is the F B g. A presentation proposing this paper and then the other curve represent Ah, brick a local search in in orange and silver lining in purple, for example. And so you see that the scaring off this purpose simulator is is rather good, and that for larger plant sizes we can get orders of magnitude faster than the state of the art approaches. Moreover, the relatively good scanning off the time to search in respect to problem size uh, they indicate that the FPD implementation would be faster than risk. Other recently proposed izing machine, such as the hope you know, natural complimented on memories distance that is very fast for small problem size in blue here, which is very fast for small problem size. But which scanning is not good on the same thing for the restricted Bosman machine. Implementing a PGA proposed by some group in Broken Recently Again, which is very fast for small parliament sizes but which canning is bad so that a dis worse than the proposed approach so that we can expect that for programs size is larger than 1000 spins. The proposed, of course, would be the faster one. Let me jump toe this other slide and another confirmation that the scheme scales well that you can find the maximum cut values off benchmark sets. The G sets better candidates that have been previously found by any other algorithms, so they are the best known could values to best of our knowledge. And, um or so which is shown in this paper table here in particular, the instances, uh, 14 and 15 of this G set can be We can find better converse than previously known, and we can find this can vary is 100 times faster than the state of the art algorithm and CP to do this which is a very common Kasich. It s not that getting this a good result on the G sets, they do not require ah, particular hard tuning of the parameters. So the tuning issuing here is very simple. It it just depends on the degree off connectivity within each graph. And so this good results on the set indicate that the proposed approach would be a good not only at solving escape problems in this problems, but all the types off graph sizing problems on Mexican province in communities. So given that the performance off the design depends on the height of this other tree, we can try to maximize the height of this other tree on a large F p g a onda and carefully routing the components within the P G A and and we can draw some projections of what type of performance we can achieve in the near future based on the, uh, implementation that we are currently working. So here you see projection for the time to solution way, then next property for solving this escape programs respect to the prime assize. And here, compared to different with such publicizing machines, particularly the digital. And, you know, 42 is shown in the green here, the green line without that's and, uh and we should two different, uh, hypothesis for this productions either that the time to solution scales as exponential off n or that the time of social skills as expression of square root off. So it seems, according to the data, that time solution scares more as an expression of square root of and also we can be sure on this and this production show that we probably can solve prime escape problem of science 2000 spins, uh, to find the rial ground state of this problem with 99 success ability in about 10 seconds, which is much faster than all the other proposed approaches. So one of the future plans for this current is in machine simulator. So the first thing is that we would like to make dissimulation closer to the rial, uh, GOP oh, optical system in particular for a first step to get closer to the system of a measurement back. See, I am. And to do this what is, uh, simulate Herbal on the p a is this quantum, uh, condoms Goshen model that is proposed described in this paper and proposed by people in the in the Entity group. And so the idea of this model is that instead of having the very simple or these and have shown previously, it includes paired all these that take into account on me the mean off the awesome leverage off the, uh, European face component, but also their violence s so that we can take into account more quantum effects off the g o p. O, such as the squeezing. And then we plan toe, make the simulator open access for the members to run their instances on the system. There will be a first version in September that will be just based on the simple common line access for the simulator and in which will have just a classic or approximation of the system. We don't know Sturm, binary weights and museum in term, but then will propose a second version that would extend the current arising machine to Iraq off F p g. A, in which we will add the more refined models truncated, ignoring the bottom Goshen model they just talked about on the support in which he valued waits for the rising problems and support the cement. So we will announce later when this is available and and far right is working >>hard comes from Universal down today in physics department, and I'd like to thank the organizers for their kind invitation to participate in this very interesting and promising workshop. Also like to say that I look forward to collaborations with with a file lab and Yoshi and collaborators on the topics of this world. So today I'll briefly talk about our attempt to understand the fundamental limits off another continues time computing, at least from the point off you off bullion satisfy ability, problem solving, using ordinary differential equations. But I think the issues that we raise, um, during this occasion actually apply to other other approaches on a log approaches as well and into other problems as well. I think everyone here knows what Dorien satisfy ability. Problems are, um, you have boolean variables. You have em clauses. Each of disjunction of collaterals literally is a variable, or it's, uh, negation. And the goal is to find an assignment to the variable, such that order clauses are true. This is a decision type problem from the MP class, which means you can checking polynomial time for satisfy ability off any assignment. And the three set is empty, complete with K three a larger, which means an efficient trees. That's over, uh, implies an efficient source for all the problems in the empty class, because all the problems in the empty class can be reduced in Polian on real time to reset. As a matter of fact, you can reduce the NP complete problems into each other. You can go from three set to set backing or two maximum dependent set, which is a set packing in graph theoretic notions or terms toe the icing graphs. A problem decision version. This is useful, and you're comparing different approaches, working on different kinds of problems when not all the closest can be satisfied. You're looking at the accusation version offset, uh called Max Set. And the goal here is to find assignment that satisfies the maximum number of clauses. And this is from the NPR class. In terms of applications. If we had inefficient sets over or np complete problems over, it was literally, positively influenced. Thousands off problems and applications in industry and and science. I'm not going to read this, but this this, of course, gives a strong motivation toe work on this kind of problems. Now our approach to set solving involves embedding the problem in a continuous space, and you use all the east to do that. So instead of working zeros and ones, we work with minus one across once, and we allow the corresponding variables toe change continuously between the two bounds. We formulate the problem with the help of a close metrics. If if a if a close, uh, does not contain a variable or its negation. The corresponding matrix element is zero. If it contains the variable in positive, for which one contains the variable in a gated for Mitt's negative one, and then we use this to formulate this products caused quote, close violation functions one for every clause, Uh, which really, continuously between zero and one. And they're zero if and only if the clause itself is true. Uh, then we form the define in order to define a dynamic such dynamics in this and dimensional hyper cube where the search happens and if they exist, solutions. They're sitting in some of the corners of this hyper cube. So we define this, uh, energy potential or landscape function shown here in a way that this is zero if and only if all the clauses all the kmc zero or the clauses off satisfied keeping these auxiliary variables a EMS always positive. And therefore, what you do here is a dynamics that is a essentially ingredient descend on this potential energy landscape. If you were to keep all the M's constant that it would get stuck in some local minimum. However, what we do here is we couple it with the dynamics we cooperated the clothes violation functions as shown here. And if he didn't have this am here just just the chaos. For example, you have essentially what case you have positive feedback. You have increasing variable. Uh, but in that case, you still get stuck would still behave will still find. So she is better than the constant version but still would get stuck only when you put here this a m which makes the dynamics in in this variable exponential like uh, only then it keeps searching until he finds a solution on deer is a reason for that. I'm not going toe talk about here, but essentially boils down toe performing a Grady and descend on a globally time barren landscape. And this is what works. Now I'm gonna talk about good or bad and maybe the ugly. Uh, this is, uh, this is What's good is that it's a hyperbolic dynamical system, which means that if you take any domain in the search space that doesn't have a solution in it or any socially than the number of trajectories in it decays exponentially quickly. And the decay rate is a characteristic in variant characteristic off the dynamics itself. Dynamical systems called the escape right the inverse off that is the time scale in which you find solutions by this by this dynamical system, and you can see here some song trajectories that are Kelty because it's it's no linear, but it's transient, chaotic. Give their sources, of course, because eventually knowledge to the solution. Now, in terms of performance here, what you show for a bunch off, um, constraint densities defined by M overran the ratio between closes toe variables for random, said Problems is random. Chris had problems, and they as its function off n And we look at money toward the wartime, the wall clock time and it behaves quite value behaves Azat party nominally until you actually he to reach the set on set transition where the hardest problems are found. But what's more interesting is if you monitor the continuous time t the performance in terms off the A narrow, continuous Time t because that seems to be a polynomial. And the way we show that is, we consider, uh, random case that random three set for a fixed constraint density Onda. We hear what you show here. Is that the right of the trash hold that it's really hard and, uh, the money through the fraction of problems that we have not been able to solve it. We select thousands of problems at that constraint ratio and resolve them without algorithm, and we monitor the fractional problems that have not yet been solved by continuous 90. And this, as you see these decays exponentially different. Educate rates for different system sizes, and in this spot shows that is dedicated behaves polynomial, or actually as a power law. So if you combine these two, you find that the time needed to solve all problems except maybe appear traction off them scales foreign or merely with the problem size. So you have paranormal, continuous time complexity. And this is also true for other types of very hard constraints and sexual problems such as exact cover, because you can always transform them into three set as we discussed before, Ramsey coloring and and on these problems, even algorithms like survey propagation will will fail. But this doesn't mean that P equals NP because what you have first of all, if you were toe implement these equations in a device whose behavior is described by these, uh, the keys. Then, of course, T the continue style variable becomes a physical work off. Time on that will be polynomial is scaling, but you have another other variables. Oxidative variables, which structured in an exponential manner. So if they represent currents or voltages in your realization and it would be an exponential cost Al Qaeda. But this is some kind of trade between time and energy, while I know how toe generate energy or I don't know how to generate time. But I know how to generate energy so it could use for it. But there's other issues as well, especially if you're trying toe do this son and digital machine but also happens. Problems happen appear. Other problems appear on in physical devices as well as we discuss later. So if you implement this in GPU, you can. Then you can get in order off to magnitude. Speed up. And you can also modify this to solve Max sad problems. Uh, quite efficiently. You are competitive with the best heuristic solvers. This is a weather problems. In 2016 Max set competition eso so this this is this is definitely this seems like a good approach, but there's off course interesting limitations, I would say interesting, because it kind of makes you think about what it means and how you can exploit this thes observations in understanding better on a low continues time complexity. If you monitored the discrete number the number of discrete steps. Don't buy the room, Dakota integrator. When you solve this on a digital machine, you're using some kind of integrator. Um and you're using the same approach. But now you measure the number off problems you haven't sold by given number of this kid, uh, steps taken by the integrator. You find out you have exponential, discrete time, complexity and, of course, thistles. A problem. And if you look closely, what happens even though the analog mathematical trajectory, that's the record here. If you monitor what happens in discrete time, uh, the integrator frustrates very little. So this is like, you know, third or for the disposition, but fluctuates like crazy. So it really is like the intervention frees us out. And this is because of the phenomenon of stiffness that are I'll talk a little bit a more about little bit layer eso. >>You know, it might look >>like an integration issue on digital machines that you could improve and could definitely improve. But actually issues bigger than that. It's It's deeper than that, because on a digital machine there is no time energy conversion. So the outside variables are efficiently representing a digital machine. So there's no exponential fluctuating current of wattage in your computer when you do this. Eso If it is not equal NP then the exponential time, complexity or exponential costs complexity has to hit you somewhere. And this is how um, but, you know, one would be tempted to think maybe this wouldn't be an issue in a analog device, and to some extent is true on our devices can be ordered to maintain faster, but they also suffer from their own problems because he not gonna be affect. That classes soldiers as well. So, indeed, if you look at other systems like Mirandizing machine measurement feedback, probably talk on the grass or selected networks. They're all hinge on some kind off our ability to control your variables in arbitrary, high precision and a certain networks you want toe read out across frequencies in case off CM's. You required identical and program because which is hard to keep, and they kind of fluctuate away from one another, shift away from one another. And if you control that, of course that you can control the performance. So actually one can ask if whether or not this is a universal bottleneck and it seems so aside, I will argue next. Um, we can recall a fundamental result by by showing harder in reaction Target from 1978. Who says that it's a purely computer science proof that if you are able toe, compute the addition multiplication division off riel variables with infinite precision, then you could solve any complete problems in polynomial time. It doesn't actually proposals all where he just chose mathematically that this would be the case. Now, of course, in Real warned, you have also precision. So the next question is, how does that affect the competition about problems? This is what you're after. Lots of precision means information also, or entropy production. Eso what you're really looking at the relationship between hardness and cost of computing off a problem. Uh, and according to Sean Hagar, there's this left branch which in principle could be polynomial time. But the question whether or not this is achievable that is not achievable, but something more cheerful. That's on the right hand side. There's always going to be some information loss, so mental degeneration that could keep you away from possibly from point normal time. So this is what we like to understand, and this information laws the source off. This is not just always I will argue, uh, in any physical system, but it's also off algorithm nature, so that is a questionable area or approach. But China gets results. Security theoretical. No, actual solar is proposed. So we can ask, you know, just theoretically get out off. Curiosity would in principle be such soldiers because it is not proposing a soldier with such properties. In principle, if if you want to look mathematically precisely what the solar does would have the right properties on, I argue. Yes, I don't have a mathematical proof, but I have some arguments that that would be the case. And this is the case for actually our city there solver that if you could calculate its trajectory in a loss this way, then it would be, uh, would solve epic complete problems in polynomial continuous time. Now, as a matter of fact, this a bit more difficult question, because time in all these can be re scared however you want. So what? Burns says that you actually have to measure the length of the trajectory, which is a new variant off the dynamical system or property dynamical system, not off its parameters ization. And we did that. So Suba Corral, my student did that first, improving on the stiffness off the problem off the integrations, using implicit solvers and some smart tricks such that you actually are closer to the actual trajectory and using the same approach. You know what fraction off problems you can solve? We did not give the length of the trajectory. You find that it is putting on nearly scaling the problem sites we have putting on your skin complexity. That means that our solar is both Polly length and, as it is, defined it also poorly time analog solver. But if you look at as a discreet algorithm, if you measure the discrete steps on a digital machine, it is an exponential solver. And the reason is because off all these stiffness, every integrator has tow truck it digitizing truncate the equations, and what it has to do is to keep the integration between the so called stability region for for that scheme, and you have to keep this product within a grimace of Jacoby in and the step size read in this region. If you use explicit methods. You want to stay within this region? Uh, but what happens that some off the Eigen values grow fast for Steve problems, and then you're you're forced to reduce that t so the product stays in this bonded domain, which means that now you have to you're forced to take smaller and smaller times, So you're you're freezing out the integration and what I will show you. That's the case. Now you can move to increase its soldiers, which is which is a tree. In this case, you have to make domain is actually on the outside. But what happens in this case is some of the Eigen values of the Jacobean, also, for six systems, start to move to zero. As they're moving to zero, they're going to enter this instability region, so your soul is going to try to keep it out, so it's going to increase the data T. But if you increase that to increase the truncation hours, so you get randomized, uh, in the large search space, so it's it's really not, uh, not going to work out. Now, one can sort off introduce a theory or language to discuss computational and are computational complexity, using the language from dynamical systems theory. But basically I I don't have time to go into this, but you have for heart problems. Security object the chaotic satellite Ouch! In the middle of the search space somewhere, and that dictates how the dynamics happens and variant properties off the dynamics. Of course, off that saddle is what the targets performance and many things, so a new, important measure that we find that it's also helpful in describing thesis. Another complexity is the so called called Makarov, or metric entropy and basically what this does in an intuitive A eyes, uh, to describe the rate at which the uncertainty containing the insignificant digits off a trajectory in the back, the flow towards the significant ones as you lose information because off arrows being, uh grown or are developed in tow. Larger errors in an exponential at an exponential rate because you have positively up north spawning. But this is an in variant property. It's the property of the set of all. This is not how you compute them, and it's really the interesting create off accuracy philosopher dynamical system. A zay said that you have in such a high dimensional that I'm consistent were positive and negatively upon of exponents. Aziz Many The total is the dimension of space and user dimension, the number off unstable manifold dimensions and as Saddam was stable, manifold direction. And there's an interesting and I think, important passion, equality, equality called the passion, equality that connect the information theoretic aspect the rate off information loss with the geometric rate of which trajectory separate minus kappa, which is the escape rate that I already talked about. Now one can actually prove a simple theorems like back off the envelope calculation. The idea here is that you know the rate at which the largest rated, which closely started trajectory separate from one another. So now you can say that, uh, that is fine, as long as my trajectory finds the solution before the projective separate too quickly. In that case, I can have the hope that if I start from some region off the face base, several close early started trajectories, they kind of go into the same solution orphaned and and that's that's That's this upper bound of this limit, and it is really showing that it has to be. It's an exponentially small number. What? It depends on the end dependence off the exponents right here, which combines information loss rate and the social time performance. So these, if this exponents here or that has a large independence or river linear independence, then you then you really have to start, uh, trajectories exponentially closer to one another in orderto end up in the same order. So this is sort off like the direction that you're going in tow, and this formulation is applicable toe all dynamical systems, uh, deterministic dynamical systems. And I think we can We can expand this further because, uh, there is, ah, way off getting the expression for the escaped rate in terms off n the number of variables from cycle expansions that I don't have time to talk about. What? It's kind of like a program that you can try toe pursuit, and this is it. So the conclusions I think of self explanatory I think there is a lot of future in in, uh, in an allo. Continue start computing. Um, they can be efficient by orders of magnitude and digital ones in solving empty heart problems because, first of all, many of the systems you like the phone line and bottleneck. There's parallelism involved, and and you can also have a large spectrum or continues time, time dynamical algorithms than discrete ones. And you know. But we also have to be mindful off. What are the possibility of what are the limits? And 11 open question is very important. Open question is, you know, what are these limits? Is there some kind off no go theory? And that tells you that you can never perform better than this limit or that limit? And I think that's that's the exciting part toe to derive thes thes this levian 10.

Hi, I'm Hideo Mabuchi from Stanford University. This is my presentation on coherent nonlinear dynamics, and combinatorial optimization. This is going to be a talk, to introduce an approach, we are taking to the analysis, of the performance of Coherent Ising Machines. So let me start with a brief introduction, to ising optimization. The ising model, represents a set of interacting magnetic moments or spins, with total energy given by the expression, shown at the bottom left of the slide. Here the cigna variables are meant to take binary values. The matrix element jij, represents the interaction, strength and sign, between any pair of spins ij, and hi represents a possible local magnetic field, acting on each thing. The ising ground state problem, is defined in an assignment of binary spin values, that achieves the lowest possible value of total energy. And an instance of the easing problem, is specified by given numerical values, for the matrix j and vector h, although the ising model originates in physics, we understand the ground state problem, to correspond to what would be called, quadratic binary optimization, in the field of operations research. And in fact, in terms of computational complexity theory, it can be established that the, ising ground state problem is NP complete. Qualitatively speaking, this makes the ising problem, a representative sort of hard optimization problem, for which it is expected, that the runtime required by any computational algorithm, to find exact solutions, should asyntonically scale, exponentially with the number of spins, and four worst case instances at each end. Of course, there's no reason to believe that, the problem instances that actually arise, in practical optimization scenarios, are going to be worst case instances. And it's also not generally the case, in practical optimization scenarios, that we demand absolute optimum solutions. Usually we're more interested in, just getting the best solution we can, within an affordable cost, where costs may be measured in terms of time, service fees and or energy required for computation. This focus is great interest on, so-called heuristic algorithms, for the ising problem and other NP complete problems, which generally get very good, but not guaranteed optimum solutions, and run much faster than algorithms, that are designed to find absolute Optima. To get some feeling for present day numbers, we can consider the famous traveling salesman problem, for which extensive compilations, of benchmarking data may be found online. A recent study found that, the best known TSP solver required median runtimes, across a library of problem instances, that scaled as a very steep route exponential, for an up to approximately 4,500. This gives some indication of the change, in runtime scaling for generic, as opposed to worst case problem instances. Some of the instances considered in this study, were taken from a public library of TSPs, derived from real world VLSI design data. This VLSI TSP library, includes instances within ranging from 131 to 744,710, instances from this library within between 6,880 and 13,584, were first solved just a few years ago, in 2017 requiring days of runtime, and a 48 core two gigahertz cluster, all instances with n greater than or equal to 14,233, remain unsolved exactly by any means. Approximate solutions however, have been found by heuristic methods, for all instances in the VLSI TSP library, with, for example, a solution within 0.014% of a known lower bound, having been discovered for an instance, with n equal 19,289, requiring approximately two days of runtime, on a single quarter at 2.4 gigahertz. Now, if we simple-minded the extrapolate, the route exponential scaling, from the study yet to n equal 4,500, we might expect that an exact solver, would require something more like a year of runtime, on the 48 core cluster, used for the n equals 13,580 for instance, which shows how much, a very small concession on the quality of the solution, makes it possible to tackle much larger instances, with much lower costs, at the extreme end, the largest TSP ever solved exactly has n equal 85,900. This is an instance derived from 1980s VLSI design, and this required 136 CPU years of computation, normalized to a single core, 2.4 gigahertz. But the 20 fold larger, so-called world TSP benchmark instance, with n equals 1,904,711, has been solved approximately, with an optimality gap bounded below 0.0474%. Coming back to the general practical concerns, of applied optimization. We may note that a recent meta study, analyze the performance of no fewer than, 37 heuristic algorithms for MaxCut, and quadratic binary optimization problems. And find the performance... Sorry, and found that a different heuristics, work best for different problem instances, selected from a large scale heterogeneous test bed, with some evidence, the cryptic structure, in terms of what types of problem instances, were best solved by any given heuristic. Indeed, there are reasons to believe, that these results for MaxCut, and quadratic binary optimization, reflect to general principle, of a performance complementarity, among heuristic optimization algorithms, and the practice of solving hard optimization problems. There thus arises the critical pre processing issue, of trying to guess, which of a number of available, good heuristic algorithms should be chosen, to tackle a given problem instance. Assuming that any one of them, would incur high cost to run, on a large problem of incidents, making an astute choice of heuristic, is a crucial part of maximizing overall performance. Unfortunately, we still have very little conceptual insight, about what makes a specific problem instance, good or bad for any given heuristic optimization algorithm. This is certainly pinpointed by researchers in the field, as a circumstance and must be addressed. So adding this all up, we see that a critical frontier, for cutting edge academic research involves, both the development of novel heuristic algorithms, that deliver better performance with lower costs, on classes of problem instances, that are underserved by existing approaches, as well as fundamental research, to provide deep conceptual insight, into what makes a given problem instance, easy or hard for such algorithms. In fact, these days, as we talk about the end of Moore's law, and speculate about a so-called second quantum revolution, it's natural to talk not only about novel algorithms, for conventional CPUs, but also about highly customized, special purpose hardware architectures, on which we may run entirely unconventional algorithms, for common tutorial optimizations, such as ising problem. So against that backdrop, I'd like to use my remaining time, to introduce our work on, analysis of coherent using machine architectures, and associated optimization algorithms. Ising machines in general, are a novel class of information processing architectures, for solving combinatorial optimization problems, by embedding them in the dynamics, of analog, physical, or a cyber-physical systems. In contrast to both more traditional engineering approaches, that build ising machines using conventional electronics, and more radical proposals, that would require large scale quantum entanglement the emerging paradigm of coherent ising machines, leverages coherent nominal dynamics, in photonic or optical electronic platforms, to enable near term construction, of large scale prototypes, that leverage posting as information dynamics. The general structure of current of current CIM systems, as shown in the figure on the right, the role of the easing spins, is played by a train of optical pulses, circulating around a fiber optical storage ring, that beam splitter inserted in the ring, is used to periodically sample, the amplitude of every optical pulse. And the measurement results, are continually read into an FPGA, which uses then to compute perturbations, to be applied to each pulse, by a synchronized optical injections. These perturbations are engineered to implement, the spin-spin coupling and local magnetic field terms, of the ising hamiltonian, corresponding to a linear part of the CIM dynamics. Asynchronously pumped parametric amplifier, denoted here as PPL and wave guide, adds a crucial nonlinear component, to the CIM dynamics as well. And the basic CIM algorithm, the pump power starts very low, and is gradually increased, at low pump powers, the amplitudes of the easing spin pulses, behave as continuous complex variables, whose real parts which can be positive or negative, by the role of soft or perhaps mean field spins. Once the pump power crosses the threshold, for perimetric self oscillation in the optical fiber ring, however, the amplitudes of the easing spin pulses, become effectively quantized into binary values, while the pump power is being ramped up, the FPGA subsystem continuously applies, its measurement based feedback implementation, of the using hamiltonian terms. The interplay of the linearized easing dynamics, implemented by the FPGA , and the thresholds quantization dynamics, provided by the sink pumped parametric amplifier, result in a final state, of the optical plus amplitudes, at the end of the pump ramp, that can be read as a binary strain, giving a proposed solution, of the ising ground state problem. This method of solving ising problems, seems quite different from a conventional algorithm, that runs entirely on a digital computer. As a crucial aspect, of the computation is performed physically, by the analog continuous coherent nonlinear dynamics, of the optical degrees of freedom, in our efforts to analyze CA and performance. We have therefore turn to dynamical systems theory. Namely a study of bifurcations, the evolution of critical points, and typologies of heteroclitic orbits, and basins of attraction. We conjecture that such analysis, can provide fundamental insight, into what makes certain optimization instances, hard or easy for coherent ising machines, and hope that our approach, can lead to both improvements of the course CIM algorithm, and the pre processing rubric, for rapidly assessing the CIM insuibility of the instances. To provide a bit of intuition about how this all works. It may help to consider the threshold dynamics, of just one or two optical parametric oscillators, in the CIM architecture just described. We can think of each of the pulse time slots, circulating around the fiber ring, as are presenting an independent OPO. We can think of a single OPO degree of freedom, as a single resonant optical mode, that experiences linear dissipation, due to coupling loss, and gain in a pump near crystal, as shown in the diagram on the upper left of the slide, as the pump power is increased from zero. As in the CIM algorithm, the non-linear gain is initially too low, to overcome linear dissipation. And the OPO field remains in a near vacuum state, at a critical threshold value, gain equals dissipation, and the OPO undergoes a sort of lasing transition. And the steady States of the OPO, above this threshold are essentially coherent States. There are actually two possible values, of the OPO coherent amplitude, and any given above threshold pump power, which are equal in magnitude, but opposite in phase, when the OPO cross this threshold, it basically chooses one of the two possible phases, randomly, resulting in the generation, of a single bit of information. If we consider two uncoupled OPOs, as shown in the upper right diagram, pumped at exactly the same power at all times, then as the pump power is increased through threshold, each OPO will independently choose a phase, and thus two random bits are generated, for any number of uncoupled OPOs, the threshold power per OPOs is unchanged, from the single OPO case. Now, however, consider a scenario, in which the two appeals are coupled to each other, by a mutual injection of their out coupled fields, as shown in the diagram on the lower right. One can imagine that, depending on the sign of the coupling parameter alpha, when one OPO is lasing, it will inject a perturbation into the other, that may interfere either constructively or destructively, with the field that it is trying to generate, via its own lasing process. As a result, when can easily show that for alpha positive, there's an effective ferromagnetic coupling, between the two OPO fields, and their collective oscillation threshold, is lowered from that of the independent OPO case, but only for the two collective oscillation modes, in which the two OPO phases are the same. For alpha negative, the collective oscillation threshold, is lowered only for the configurations, in which the OPO phases are opposite. So then looking at how alpha is related to the jij matrix, of the ising spin coupling hamilitonian, it follows the, we could use this simplistic to OPO CIM, to solve the ground state problem, of the ferromagnetic or antiferromagnetic angles, to ising model, simply by increasing the pump power, from zero and observing what phase relation occurs, as the two appeals first start to lase. Clearly we can imagine generalizing the story to larger, and, however, the story doesn't stay as clean and simple, for all larger problem instances. And to find a more complicated example, we only need to go to n equals four, for some choices of jij for n equals four, the story remains simple, like the n equals two case. The figure on the upper left of this slide, shows the energy of various critical points, for a non frustrated n equals for instance, in which the first bifurcated critical point, that is the one that, by forgets of the lowest pump value a, this first bifurcated critical point, flows asyntonically into the lowest energy using solution, and the figure on the upper right, however, the first bifurcated critical point, flows to a very good, but suboptimal minimum at large pump power. The global minimum is actually given, by a distinct critical point. The first appears at a higher pump power, and is not needed radically connected to the origin. The basic CIM algorithm, is this not able to find this global minimum, such non-ideal behavior, seems to become more common at margin end, for the n equals 20 instance show in the lower plots, where the lower right pod is just a zoom into, a region of the lower left block. It can be seen that the global minimum, corresponds to a critical point, that first appears that of pump parameter a around 0.16, at some distance from the adriatic trajectory of the origin. That's curious to note that, in both of these small and examples, however, the critical point corresponding to the global minimum, appears relatively close, to the adiabatic trajectory of the origin, as compared to the most of the other, local minimum that appear. We're currently working to characterise, the face portrait typology, between the global minimum, and the adiabatic trajectory of the origin, taking clues as to how the basic CIM algorithm, could be generalized, to search for non-adiabatic trajectories, that jumped to the global minimum, during the pump up, of course, n equals 20 is still too small, to be of interest for practical optimization applications. But the advantage of beginning, with the study of small instances, is that we're able to reliably to determine, their global minima, and to see how they relate to the idea, that trajectory of the origin, and the basic CIM algorithm. And the small land limit, We can also analyze, for the quantum mechanical models of CAM dynamics, but that's a topic for future talks. Existing large-scale prototypes, are pushing into the range of, n equals, 10 to the four, 10 to the five, 10 to the six. So our ultimate objective in theoretical analysis, really has to be, to try to say something about CAM dynamics, and regime of much larger in. Our initial approach to characterizing CAM behavior, in the large end regime, relies on the use of random matrix theory. And this connects to prior research on spin classes, SK models, and the tap equations, et cetera, at present we're focusing on, statistical characterization, of the CIM gradient descent landscape, including the evolution of critical points, And their value spectra, as the pump powers gradually increase. We're investigating, for example, whether there could be some way, to explain differences in the relative stability, of the global minimum versus other local minima. We're also working to understand the deleterious, or potentially beneficial effects, of non-ideologies such as asymmetry, in the implemented using couplings, looking one step ahead, we plan to move next into the direction, of considering more realistic classes of problem instances, such as quadratic binary optimization with constraints. So in closing I should acknowledge, people who did the hard work, on these things that I've shown. So my group, including graduate students, Edwin Ng, Daniel Wennberg, Ryatatsu Yanagimoto, and Atsushi Yamamura have been working, in close collaboration with, Surya Ganguli, Marty Fejer and Amir Safavi-Naeini. All of us within the department of applied physics, at Stanford university and also in collaboration with Yoshihisa Yamamoto, over at NTT-PHI research labs. And I should acknowledge funding support, from the NSF by the Coherent Ising Machines, expedition in computing, also from NTT-PHI research labs, army research office, and ExxonMobil. That's it. Thanks very much.

>> Announcer: Live, from Orlando, Florida, it's the Cube covering Grace Hopper's celebration of women in computing brought to you by Silicon Angle Media >> Welcome back to the Cube's coverage of the Grace Hopper conference here in Orlando, Florida I'm your host Rebecca Knight, along with my co host Jeff Frick. We're joined by Megan Smith. We're very excited to have you on the show. >> It's good to be here >> She is the third US CTO and also the CEO of a new company, Shift7.co, so thanks so much for joining us. >> Thanks for having me, it's great to be here. It's so fun to be at Hopper, >> Rebecca: It is, it is! >> It's cool, it's the Grace Hopper celebration, because we're trying to celebrate women in computing, and we're what, at 18 thousand people now, >> The biggest ever, >> Plus I think, 6 thousand people joining on the livestream, which is great. >> Before the cameras were rolling, we were talking about your role as the 3rd US CTO, and just talking about getting more technology into government to help leaders work together, and move faster. Tell us a little about this initiative. >> What's so great, is it's not partisan, fixing the government and making it work better, so all the work that we were doing continues. What we were able to put in place, during the Obama administration, and continues to Trump, were things like, the CT office. Having technical people, so I worked at Google, people work at Amazon, Facebook, Twitter, these companies who have that background, to join in on policy conversations, one, to join in on capacity building the government, so data sciences and tech and, let's have our services be as great as Amazon, or as Twitter, or Oracle, and not be sort of retro, really serve the American people. And then also, helping the American people in general, with capacity building, things like computer science for all. So that was an initiative that continues to get all of our children to have coding at school. That all children, you couldn't graduate from high school without having had some experience on learning of coding Coding is a 21st century fluency, it's a skill we all need, Like freshman biology. You want to know some biology, you want to know some coding, you want to know how to write, so making sure they have is tech-up, which was a program we started to help train Americans, there's six hundred thousand jobs open, in the United States, and they pay 50% more than the average American salary. The companies are starving. How do we rapidly get more Americans into these jobs? It turns out that people have, of course, created these fabulous code boot camps, you can train in three months for these jobs, some of them are paid, some times they pay you, all different kinds, some are online, some are offline, they're all over the country. So we're able to get more people to consider, a job like that, culturally they think, Well I don't, why would I, I don't know how to do that. Well you can, this is a fun and interesting and exciting career, you can do digital marketing, you can do user interface design. You can get involved in front end or back end coding, product management, all those things, sales. And so, how do you pull lots more Americans in, get our companies fueled so we have really the economic opportunity, and they're all over the country. Location wise, and topic wise. So we did tech hour now, and a tech jobs tour, which is not what we did in government, but we continue some of that work. >> This weird dichotomy, because on one end, people are worried about tech taking jobs, on the other hand, there's a ton of open tech jobs. And there's this transition period, that's difficult, obviously for people that didn't grow up, but one of the keynote speakers today, told a really heartening story, that she didn't get into it until the day she had to leave her abusive husband, and now she is a coder >> That's Doctor Sue Black, who was just given the Order of the British Empire, I mean, she is an incredible computer scientist. Yes, she escaped an abusive marriage with three small children, in her early 20s, I think. Ended up moving into public housing, and dealing with three children only being the school from 9 until 3, and eventually getting her PhD in computer science, and really, she started Techmoms now, she continues to do research in other things, but she's really trying to use her story, and her organizing capacity, to have more people realize this isn't hard like figuring out gravity waves that won the Nobel prize. This is hard like writing a hard essay, so we all can learn to write an essay. It takes some mastery work, you don't learn it in kindergarten but by the time you're in 7th, 8th, 9th, 10th, 12th grade, you can do it. >> It's not rocket science. >> Right, so coding is like that. >> The other piece you said that's very interesting, is the consumerization of IT. We've seen it at enterprise, a huge trend. But, now I expect everything that's on my phone, when I interact with Facebook or Amazon, or whatever, to be in all the applications, so, as you said, that's influencing government, and the way they have to deliver services, and I would imagine, too, with kind of the next wave of kids coming in, graduating, going into public service, they certainly have that expectation, right? They've been working on their phone forever of course it should be on the phone. >> And so we want everybody in our country fluent in computer science and coding at a basic level, like again, like freshman biology or takin' chemistry in high school, or taking writing. So that everyone could realize this is not rocket science we could have these kinds of capabilities as part of our services, from Housing and Urban Development, from the Department of Education. You know, a lot of us use our phones to get places, you know, on our maps, and so that's actually data coming from the US Geological survey, if you're looking at the weather, you're looking at NOAA's satellites, this is open government data. We were able to open over two hundred thousand data sets, from all over government, not private data, but public data, that you could make an entire app store, or Google play set of products on top of that. Government wouldn't have to pay for that, it just packages up the API as well. A really good example of that, is the US census team. There's nothing more big data than census, they have all of our information from a data perspective, and so they did opportunity.census.gov, and they said to various agencies, let us help you bridge these data sets into something that someone could build on top of, like we're seeing from the courses sector, we saw wonderful things like, Housing and Urban Development said, okay, our challenges are housing affordability, mobility, these are the challenges instead of having HUD make an app for Americans to come to, they just can explain what their problem is, what data sets, and then engage extraordinary companies, like airbnb, Redfins, Zillow, these fabulous tech companies, who can make instead a product for 100% of the Americans, rather than only wealthy or middle class Americans, and so they did things like, opportunity score, and airbnb helping you figuring out, if I rent a room in my house I can make my rent more affordable, very creative apps, that we can see, same thing for the Department of Ed or Department of Labor, and as the data gets out there, and as apps come, also the opportunity for data science and machine learning. What if, as much as we serve ads to ourselves, in these algorithms, what if we use the algorithms to help Americans find a job that they would love? You know, job matching, and these kinds of opportunities. of the problems in the world, and helping government get more fluent at that. And the way to do that is not so much, jam the government You have to do this, but find terrific talent like we see at Hopper, and have them cycle into the government, to be co-leaders just like a surgeon general would come. >> Are you facing recruitment challenges in that same way though? In the sense that technology is having a hard enough time recruiting and retaining women, but the government, too, is that seen as enough of an employer of choice for young talented, bright ambitious, young women? >> I'm not in government now, but when we were in there, we found a very interesting thing. Alex Mcgovern, who had been the general counsel of Twitter who was Stephanie's CTO with me and led a lot of our tech quals we called TQ like tech IQ in policy, together with economists and lawyers and others have if we're going to decide net neutrality, let's include everyone, including computer scientists, and we're going to sue bridge and open source, So we talked about that, and on the way going in Mcgovern, he said, wouldn't it be cool if, just like when you look at a lawyer's resume, you might see that they clerk and they served their county through clerking and through the judicial system, as well as being a private lawyer, they were a public defender, that's a pretty normal thing to see on a legal resume. If you looked at medical, you might see them going into NIH or doing some research, if you looked at a scientist, they might have gone to, done some NSF work or others. But for the tech crew, there is of course amazing technical people in NASA, NAH and the Department of Energy, and there's great IT teams, but there's not this thing that the Silicon Valley crew resume would say, oh, yeah, I served my country. So that's why, under President Obama, we were able to create these incredible programs. The Presidential Innovation Fellows, which was a one year term of service, The United States Digital Service, which is a three months to a two year term of service in the VA. What's more amazing if you build Amazon, than to go as a second act and help our veterans? It's an incredible honor, to the point of, will they come? Yes, that's what we were hoping, could we have that be a normal thing, and yes it's become a normal thing. And the Trump administration continues it. The 18F team is in the general services administration, they're on 18th and F so they have a code name. But that particular team is located around the country, not only in DC but in San Francisco, in Chicago, and others. So you see this tech sector flowing now into the government on a regular basis, and we welcome more peoples. The government is who shows up to help, so we need the tech sector to show up cause we've got a lot of money as a country, but if we're not effectively using it we're not serving the American people and foster children, veterans, elders, others need the services that they deserve and we have the money, so let's make it happen the way the tech sector is delivering Amazon packages or searches. >> What is your feeling, this is not your first Grace Hopper obviously, but what is your feeling about this conference, and advice that you would give to young women who are here, maybe for their first or second time, in terms of getting the most their time here? >> You know, I think the main thing is, it's a celebration, that's fun and you can walk up to anyone, so just talk to everyone. I've been talking to a million people on the floor, fabulous. Students are here, more senior technical leaders are here. We've been running speed mentoring, we're running a program called the Tech Jobs Tour, it's at Techjobstour.com, it's a #Americanshiring, and we've been going to 50 different cities and so we're running a version of that, and we do speed mentoring, so come to the speed mentoring sessions, it's a five minute pop, talk to someone about what you're tryin' to do. Lot's of programs like that, get into the sessions, come to the keynotes which are so inspiring, and Melinda Gates was amazing today, Dr. Fefe Lee was incredible, just across aboard, Dr Sue Black was here, I thought it was great today, actually, just to reflect on Melinda's keynote, I think this might have been the first time, I was talking to her, that she's really talked about her own technical experience >> That struck me, too! As a coder, starting in computer science. I didn't really understand that she had really started very early, with Apple 3 and the story of her dad >> And her love of her Apple 3, right! and really high school coding, which is so important for young people in high school and middle school, even younger. The Muscogee Creek Tribe, in Oklahoma, is teaching robotics in head start, so we can start in preschool. Just make it fun, and interesting. They're funny, they don't do battle bots, because you don't really want to teach 3 and 4 year olds to fight, so instead they have collaborative robots. >> Robots who work together Age appropriate. >> Well Megan Smith, this has been so fun talking to you, thanks so much for coming on our show. >> Thanks for having me. >> We will have more from the Grace Hopper Conference just after this, I'm Rebecca Knight for Jeff Frick (music)

>> Announcer: Live from Orlando, Florida. It's theCUBE, covering Grace Hopper's Celebration of Women in Computing, brought to you by Silicon Angle Media. (bright music) >> Welcome back to the Cube's coverage of the Grace Hopper Conference here in Orlando, Florida. I'm your host Rebecca Knight. I'm joined by Ayanna Howard. She is a professor at the Georgia Institute of Technology and also Chief Technology Officer at Zyrobotics. >> Thank you. >> Thanks so much for joining us. >> Thank you very much for having me. >> So start to tell our viewers a little bit about Zyrobotics. I know it was a spin-off of your research that you were doing at Georgia Tech. >> Yeah, so interesting enough Zyrobotics, so at Georgia Tech I focus on working in technologies, robotics for children with special needs. Primarily children with motor disabilities, cerebral palsy for example, children with autism. And so one of the things as we had developed was the ability to access computing technology because I was running robot programming camp. So I was running camps for all children, so an inclusive camp and I had typical children and children with special needs, and what happened was people kept asking me, "Oh, can we take this home?" It was like, "Yeah, no, (laughing) "that's got to stay in the lab, sorry. "But you can bring your kid back." And so the company really came out of trying to commercialize that special technology that allows inclusiveness for kids in this kind of STEM education. So that's how Zyrobotics came about. >> So talk a little bit about the technology. What does it do? How does it help kids with these different learning needs? >> So imagine you have a child who has motor limitation, and if you look now, so much is on tablets. Tablets, smartphones, even education. And if I have a motor disability, have you ever tried swiping with your fist? Right, or even if you're an older adult, and taking your finger, and if you have a tremor, like moving things around, so this is very difficult. And yet that is the way the technology is made, which isn't a service. It's just not made for everyone. And so what we've done is we've created these devices, very fun, think of it as a stuffed animal, that allows you to, if you want to stomp, if you want to do your finger, if your access point is in your foot, and you just tap your foot, it allows you to interact with the different educational apps. But what we found is that typical kids also like (laughing) playing with the toys. >> Rebecca: Right, right, right. >> So it's like, oh what is this? This is interesting. And so that's why it provided this nice blend of kids of any ability the ability to access these educational apps. So but you also are a full-time professor at Georgia Tech, and you run a traineeship in healthcare robotics. Tell our viewers a little bit more about that. >> Yeah, so I run a program called ARMS, so it's funded by the National Science Foundation. And what I've found is, a long time ago, the way that we were training our computer science students, our engineering students in robotics was typically I would say ad hoc. So I'd have a student, and they were like, "I'm interested in healthcare robotics." And I would call up my clinician friend and say, "hey, can we do an observation?" And my student would go there and basically shadow a therapist or a doctor for the day. And then they go back. And so this is what I was doing. And I found out that most professors who had students in healthcare-related activities were doing the same thing. And I was like, wait, hold it. This sounds like it's more than just me. Maybe we can formalize this a little bit more. And so the trainee-ship program actually takes roboticist students and immerses them in the medical side. And so for example this past summer, they spent the entire summer over in the clinic and the hospital watching surgeries, I mean actually scrubbing up, following patients, understanding what is Parkinson's and how do you do assessments. And so they were fully immersed as if they were medical resident students, or resident person in the clinic. And what happens is, then, and this is all in their first year, they come back into their studies, and now they understand, "okay, if I'm designing "this technology, what does it mean "if I'm designing for someone who's recovering from stroke? "What does that really mean?" And they have a vision of the patients, not just their own, I mean, they have a real vision of Mister Joe, that they've worked with and how he might have struggled with some concept and what they're doing can actually enable. And so it gives engineers, scientists, roboticists that power. >> And the empathy to really understand how it will be used. >> Yes, and understand that and not build or design in a box, which is really unfortunate that sometimes we do that. We design based on our own beliefs, not taking into account that there are other users and you are not the user, necessarily, of your own technology. >> So I want talk a little bit about this conference. This is your third Grace Hopper Conference. What does it mean to you to be here, and what do you get out of it? Are you here for Zyrobotics? Are you here for Georgia Tech? >> I am here for women in computing. And so it's actually not linked to a specific company or an organization. It's the fact that I feel a responsibility, they call me a role model, but- >> Rebecca: We're going to go with it, we're going to go with it. >> We're going to go with it. (laughing) I mean, I had a lot of mentors growing up. Not many were women. It's only at my later age that I've actually met some great, great women mentors. And so I feel a responsibility to come to Grace Hopper and just talk, share my experiences, sometimes be vulnerable and open to the trials and tribulations, but then the pure joy you get from staying in the field and the pure joy you get from actually impacting the world with your mind, with your technology, with your stuff. And I think it's amazing how, to be here and see all these young ladies, both students and older, well-established women leaders, and say, "yeah, we got this. "We can change the world with our power." >> So we're really at this inflection point in technology where problems, the biases, the barriers that have kept women from progressing, from first of all getting into the field and also progressing, are really front-page news. And sort of the problems that women have faced in the industry, the sexism, is really being talked about. But is that a good thing in the sense, I mean, yes, it's one thing to get these problems out there, but are we also discouraging women because it's showing women how tough it is to be in this industry and this bro-grammer culture? >> I think it's a two-edged sword. So in one instance, these things were happening anyway. And if you actually look at retention, which is surprising, retention of women who've been in the computing field for a longer period of time, a lot of them were dropping out. It's like, wait, hold it. You got through the pipeline, what happened? And so we all knew a lot of this stuff was going on. We have first-hand experience with it. And so the conversation now is letting everyone know about it. And I think that's how anything happens. It's that others are like, "I didn't realize." others start empathizing. "I didn't realize that this is what you were "going through. "What can I do to help?" Even if they are not necessarily a woman or a minority. And so I think what happens is by having that conversation, it makes everyone aware of it so that things can start changing. It's a negative, the fact that maybe young women are like, "oh, I don't want to go through that." I think by having role models that are like, "hey, yeah, that's what it's like, "but guess what, I'm running the company. "I'm the CEO, and so imagine what it'd be like "if you come in now that the conversation is open "versus what I was going through "when nobody was talking about it." We didn't have anyone to say, "hey, can you help me? "I just need some assistance, just to talk about something." Now you can, you can be open about it. >> So what is your advice? I mean, we know that the numbers are bleak. Tech is comprised of 25% women, 15% in leadership positions. For black and Latina, it's abysmal. What do you tell your students about this industry? >> So I tell my students, one is, if you want to change the world, and usually students that take my course and work with me are ones that want to have an impact with their minds and their technology, and so my thing is if you want to change the world, computer science, engineering is the only way that you can because the world is based on you and your technology. And in fact, if you don't, I put in the guilt, if you don't get involved in this, then the world is not going to change. And your kids' kids will have to live in this world that you have. So it's really your responsibility (laughing) to get into this space. >> The guilt is good, that's good, yeah. >> It is, for women, guilt is really good. >> I know, it's powerful, so powerful. >> Yeah, yeah. >> I want to talk a little bit about funding because I know that your trainee program, it's partly funded by the National Science Foundation. So funding is such a hot topic here, and whether you're a female entrepreneur who's trying to get money for your idea or you're a scientist trying to fund your research, tell us a little bit about the landscape, what you're seeing, what you're feeling. >> I would say that government funding, so the National Science Foundation, I would say NIH, there is more equality in the representation. >> Rebecca: There is more equality. >> It's not 50-50. But you have a fighting chance, right? I would argue, though, that in the startup world, you need to go for government funding and non-profits that may be angels because honey, VCs are not going to look at you. I truly believe that, and being a startup company, I talked to a lot of women entrepreneurs who have broke in the VC field, and they tell me basically how many frogs they had to kiss, you know? And so I think that landscape has not changed as much. But I think funding as a scientist for government grants, I think it's more, it's not fair, but it's more equal because in government, it's okay for you to say, as a program manager, "hey, something's wrong here." Because the government represents the population. So it's okay as a program manager to say that. I don't know that it's as safe to say that as a VC, like, "hey, our company portfolio doesn't look "like the rest of America." >> Right, right. So your advice there for female entrepreneurs or female researchers trying to get money is to go first to either angels or the government. >> I say that will help you keep your company alive. But you still have to kiss a lot of frogs. You still do. And eventually you will find a frog that turns into a princess and will fund you. But if you think about, how do you survive through this company and how do you keep it to the next levels, you go through any type of funding resource that you can. And so if the angel funding world in terms of government, it's not a guarantee, but it's easier, grab that, non-diluted, by the way, typically, until you go the VC direction. >> Now, in terms of the funding environment, though, NIH and NSF, do you feel they're giving as much money right now? We have an administration that is... >> Yeah, no, so overall the budgets themselves are, so NSF and NIH, this last cycle they kind of weathered a cut. But if you look overall over the last umpteen years, you see that the rate of acceptance has dropped because there's a lot more researchers going for funding, the budget doesn't keep up, necessarily, with the cost of living expenses kind of thing, cost for tuition, cost for grad students. And so overall the funding has declined. But that is not a gender issue. That is a issue just about the value of basic research in general. And the US, a lot of us understand but a lot of us do not. And so we feel that in terms of the funding process. >> So as a professor but then also as someone who's working in industry, how do you make sure that women can see themselves and see potentially rich and rewarding careers? >> So I do a couple of activities. For example, I'm going to talk about one, which CRWA grad cohort. And so what that focuses on is graduate students, women, either PhD, Master's wanting to be a PhD, and what we do is we provide those mechanisms for them to interact with community members. So we bring in these- >> Rebecca: So this is not just at Georgia Tech. This is nationwide. >> This is nationwide. Young women, they come in, like, "oh, what is this?" First off, they get to see other of their peers at other schools. Second is we bring in senior women that are doing exceptionally well, and they do things like one on one mentorship. They share. So we select these women who are open to sharing their experiences, both the good and the bad, and so it provides that network of, "okay, look, it might be hard in grad school, "but we have a peer network, take advantage. "And there are senior women you can take advantage, "to talk to and kind of ping them on different issues "that you have." So I think programs like that, and we're not the only one, but programs like CRWA grad cohort, CRAW URM, undergraduate cohort, are ways to ensure that you don't get discouraged at a younger age. >> So Zyrobotics, it's founded in 2013. What is the future of it? I mean, it's such an exciting technology and one that I think really has a lot of uses because as you said, it's not only for children but it could be for stroke victims, for aging people who are sort of losing some of their mobility. >> So my goal, I always say five years, right? So when I started it was like, five year goal cause that's like the holy grail, you make it for five years. So we're at year four, we just crossed. So we're in that five years. But what I see more as the vision, what I would say the secret magic of Zyrobotics is to make sure that accessibility is an integral part of the conversation. It's not an afterthought, it's not a someone designed technology, oh, let's think about accessibility and inclusiveness after the fact. And so I'm hoping that one, the product of course takes off, but also that it starts changing the conversation a little bit. So for example, I go out, I talk about how do you design technology that is really, really cool, is cutting edge, that's accessible at its core. It's accessible to the different learning ways, different access ways that people have of interacting with technology. How do you get that message across that, "hey, you can so this and you can still make money." So it's not like oh, accessibility, we can't make any money. Like, no, you can actually still make money even if it's a core value. So that's my vision is to have basically, have Zyrobotics lead that but then have other companies adopt it as, "oh, yeah, why haven't we done this? "Yeah, this makes total, total sense." >> Great, Ayanna Howard, thank you so much for joining us. It's been a pleasure having you on theCUBE. >> Thank you, this was fun. Thank you for the invite. >> I'm Rebecca Knight, here in Orlando, Florida at Grace Hopper. We will have more just after this. (bright music)