Princeton Robotics Seminar: Vikas Sindhwani, Friday June 9 at 11am
Speaker: Vikas Sindhwani, Senior Staff Research Scientist, Google Deepmind Date: Friday, June 9, 2023 Time: 11am ET Location: Virtual, Zoom link: [ https://princeton.zoom.us/my/robotics | https://princeton.zoom.us/my/robotics ] Event page: [ https://robo.princeton.edu/seminar | https://robo.princeton.edu/seminar ] Title: Large Language Models with Eyes, Arms and Legs Abstract: To become useful in human-centric environments, robots must demonstrate language comprehension, semantic understanding and logical reasoning capabilities working in concert with low-level physical skills. With the advent of modern "foundation models" trained on massive datasets, the algorithmic path to developing general-purpose “robot brains” is (arguably) becoming clearer, though many challenges remain. In the first part of this talk, I will attempt to give a flavor of how state-of-the-art multimodal foundation models are built, and how they can be bridged with low-level control. In the second part of the talk, I will summarize a few surprising lessons on control synthesis observed while solving a collection of Robotics benchmarks at Google. I will end with some emerging open problems and opportunities at the intersection of dynamics, control and foundation models. Bio: Vikas Sindhwani is Research Scientist at Google Deepmind in New York where he leads a research group focused on solving a range of planning, perception, learning, and control problems arising in Robotics. His interests are broadly in core mathematical foundations of statistical machine learning, and in end-to-end design aspects of building large-scale and robust AI systems. He received the best paper award at Uncertainty in Artificial Intelligence (UAI-2013), the IBM Pat Goldberg Memorial Award in 2014, and was finalist for Outstanding Planning Paper Award at ICRA-2022. He serves on the editorial board of Transactions on Machine Learning Research (TMLR) and IEEE Transactions on Pattern Analysis and Machine Intelligence; he has been area chair and senior program committee member for NeurIPS, International Conference on Learning Representations (ICLR) and Knowledge Discovery and Data Mining (KDD). He previously headed the Machine Learning group at IBM Research, NY. He has a PhD in Computer Science from the University of Chicago and a B.Tech in Engineering Physics from Indian Institute of Technology (IIT) Mumbai. His publications are available at: http://vikas.sindhwani.org/
Speaker: Manya Ghobadi, from Massachusetts Institute of Technology Date: Monday, July 31, 2023 Time: 2pm ET Location: CS 105 Host: Jennifer Rexford Event page: [ https://www.cs.princeton.edu/events/26450 | https://www.cs.princeton.edu/events/26450 ] Title: Next-Generation Optical Networks for Machine Learning Jobs Abstract: In this talk, I will explore three elements of designing next-generation machine learning systems: congestion control, network topology, and computation frequency. I will show that fair sharing, the holy grail of congestion control algorithms, is not necessarily desirable for deep neural network training clusters. Then I will introduce a new optical fabric that optimally combines network topology and parallelization strategies for machine learning training clusters. Finally, I will demonstrate the benefits of leveraging photonic computing systems for real-time, energy-efficient inference via analog computing. I will discuss that pushing the frontiers of optical networks for machine learning workloads will enable us to fully harness the potential of deep neural networks and achieve improved performance and scalability. Bio: Manya Ghobadi is faculty in the EECS department at MIT. Her research spans different areas in computer networks, focusing on optical reconfigurable networks, networks for machine learning, and high-performance cloud infrastructure. Her work has been recognized by the ACM-W Rising Star award, Sloan Fellowship in Computer Science, ACM SIGCOMM Rising Star award, NSF CAREER award, Optica Simmons Memorial Speakership award, best paper award at the Machine Learning Systems (MLSys) conference, as well as the best dataset and best paper awards at the ACM Internet Measurement Conference (IMC). Manya received her Ph.D. from the University of Toronto and spent a few years at Microsoft Research and Google before joining MIT. For questions about this event, please contact Sophia Yoo at [ mailto:sy6@princeton.edu | sy6@princeton.edu ] .
Speaker: Manya Ghobadi, from Massachusetts Institute of Technology Date: Monday, July 31, 2023 Time: 2pm ET Location: CS 105 Host: Jennifer Rexford Event page: [ https://www.cs.princeton.edu/events/26450 | https://www.cs.princeton.edu/events/26450 ] To attend remotely via webinar, please register here: [ https://princeton.zoom.us/webinar/register/WN_LyskDcd3QDyJJwGuqNr46g | https://princeton.zoom.us/webinar/register/WN_LyskDcd3QDyJJwGuqNr46g ] Title: Next-Generation Optical Networks for Machine Learning Jobs Abstract: In this talk, I will explore three elements of designing next-generation machine learning systems: congestion control, network topology, and computation frequency. I will show that fair sharing, the holy grail of congestion control algorithms, is not necessarily desirable for deep neural network training clusters. Then I will introduce a new optical fabric that optimally combines network topology and parallelization strategies for machine learning training clusters. Finally, I will demonstrate the benefits of leveraging photonic computing systems for real-time, energy-efficient inference via analog computing. I will discuss that pushing the frontiers of optical networks for machine learning workloads will enable us to fully harness the potential of deep neural networks and achieve improved performance and scalability. Bio: Manya Ghobadi is faculty in the EECS department at MIT. Her research spans different areas in computer networks, focusing on optical reconfigurable networks, networks for machine learning, and high-performance cloud infrastructure. Her work has been recognized by the ACM-W Rising Star award, Sloan Fellowship in Computer Science, ACM SIGCOMM Rising Star award, NSF CAREER award, Optica Simmons Memorial Speakership award, best paper award at the Machine Learning Systems (MLSys) conference, as well as the best dataset and best paper awards at the ACM Internet Measurement Conference (IMC). Manya received her Ph.D. from the University of Toronto and spent a few years at Microsoft Research and Google before joining MIT. For questions about this event, please contact Sophia Yoo at [ mailto:sy6@princeton.edu | sy6@princeton.edu ] .
Speaker: Sagi Snir from Technion Israel Date: Thursday, August 17 Time: 12:30pm Location: CS 301 Host: Ben Raphael Title: Horizontal Gene Transfer Phylogenetics: A Random Walk Approach Abstract. The dramatic decrease in time and cost for generating genetic sequence data has opened up vast opportunities in molecular systematics, one of which is the ability to decipher the evolutionary history of strains of a species. Under this fine systematic resolution, the standard markers are often too crude to provide a reliable phylogenetic signal. Nevertheless, among prokaryotes, genome dynamics (GD) in the form of horizontal gene transfer (HGT), the transfer of genetic material between organisms not through lineal descent, seem to provide far richer information by affecting both gene order and gene content. The synteny index (SI) between a pair of genomes combines these latter two factors, allowing comparison of genomes with unequal gene content, together with order considerations of their common genes. Although this approach is useful for classifying close relatives, no rigorous statistical modelling for it has been suggested. Such modelling is valuable, as it allows observed measures to be transformed into estimates of time periods during evolution, yielding the additivity of the SI measure. To the best of our knowledge, there is no other additivity proof for other gene order/content measures under HGT. Here we provide a first statistical, two-level modeling and analysis, for GD under a very simple operation – the Jump operation. Under this framework, at the higher level, genome evolution is modeled as a random walk in the genome permutation state space. At the lower level, gene neighborhood is modeled as a birth–death–immigration process affected by the genes jumping across the genome. Using this modeling we can infer several neutral characteristics for genomes evolving along a tree and analytically relate the HGT rate and time to the expected SI. We applied this model to the new version of the orthology DB EggNOG containing over 4.5K taxa. To the best of our knowledge, this is the largest gene-order-based tree constructed and it overcomes shortcomings found in previous approaches. Constructing a GD-based tree allows toconfirm and contrast findings based on other phylogenetic approaches, as we show. Related Papers: 1) Katriel G., Mahanaymi U., Koutschan C., Zeilberger D., Steel M. and Snir S. 2023. Gene Transfer-based Phylogenetics: Analytical Expressions and Additivity via Birth–Death Theory, Accepted at Systematic Biology. A preliminary version is available at bioarxiv 2) Sevillya, G., D. Doerr, Y. Lerner, J. Stoye, M. Steel, and S. Snir. 2019. Horizontal Gene Transfer Phylogenetics: A Random Walk Approach. Molecular Biology and Evolution (MBE)37:1470–1479. 3) Shifman, A., N. Ninyo, U. Gophna, and S. Snir. 2013. Phylo si: a new genome-wide approach for prokaryotic phylogeny. Nucleic acids research (NAR)42:2391–2404. Bio: Sagi Snir graduated in Computer Science from the Technion Israel, focusing on analytical, algebraic, maximum likelihood solutions to phylogenetics. After a postdoc in Math and Computer Science depts at UC Berkeley, he returned to the University of Haifa in Israel, where he has established the Bioinformatics program for grad students. He is now a professor of computational evolution at the University of Haifa and the President of the Israeli Society for Bioinformatics and Computational Biology. His research combines algorithmic and combinatorial/statistical approaches to problems from evolution with focus on phylogenetic trees and networks. He has developed the Quartet MaxCut algorithm to combine conflicting signals between evolutionary trees and other fundamental results on maximum likelihood of trees and networks. His papers have been published in both leading pure theoretical computer science venues and pure evolution venues.
Speaker: Vladlen Koltun Date: Tuesday, September 12, 2023 Time: 4:30pm ET Location: Friend Center, room 101 Host: Jia Deng Event page: [ https://www.cs.princeton.edu/events/26479 | https://www.cs.princeton.edu/events/26479 ] Title: A Quiet Revolution in Robotics Bio: [ http://vladlen.info/ | V ] [ http://vladlen.info/ | ladlen Koltun ] received his PhD in 2002 and has worked across multiple fields of computer science. He has mentored more than 50 PhD students, postdocs, research scientists, and PhD student interns, many of whom are now successful research leaders. Until 2021, he had served as the Chief Scientist for Intelligent Systems at Intel, where he built an international research lab, based on four continents, that produced high-impact results in robotics, computer vision, image synthesis, machine learning, and other areas. This event is co-sponsored by the Department of Computer Science and the Center for Statistics and Machine Learning and will be recorded and live streamed on Princeton University Media Central. [ https://mediacentrallive.princeton.edu/ | See link here ] .
Speaker: Vladlen Koltun Date: Tuesday, September 12, 2023 Time: 4:30pm ET Location: Friend Center, room 101 Host: Jia Deng Event page: [ https://www.cs.princeton.edu/events/26479 | https://www.cs.princeton.edu/events/26479 ] Title: A Quiet Revolution in Robotics Bio: [ http://vladlen.info/ | V ] [ http://vladlen.info/ | ladlen Koltun ] received his PhD in 2002 and has worked across multiple fields of computer science. He has mentored more than 50 PhD students, postdocs, research scientists, and PhD student interns, many of whom are now successful research leaders. Until 2021, he had served as the Chief Scientist for Intelligent Systems at Intel, where he built an international research lab, based on four continents, that produced high-impact results in robotics, computer vision, image synthesis, machine learning, and other areas. This event is co-sponsored by the Department of Computer Science and the Center for Statistics and Machine Learning and will be recorded and live streamed on Princeton University Media Central. [ https://mediacentrallive.princeton.edu/ | See link here ] .
participants (1)
-
Emily C. Lawrence