- talks - lists.cs.princeton.edu

Princeton Robotics Seminar: Friday Oct 28 at 11am
by Emily C. Lawrence 25 Oct '22

25 Oct '22

Princeton Robotics Seminar Speaker: Tomás Lozano-Pérez, Professor of Computer Science & Engineering, MIT Date: Friday Oct 28, 2022 Time: 11am EDT Location: Computer Science 105 Event page: [ https://robo.princeton.edu/seminar | https://robo.princeton.edu/seminar ] Title: Generalization in Planning and Learning for Robotic Manipulation Abstract: An enduring goal of AI and robotics has been to build a robot capable of robustly performing a wide variety of tasks in a wide variety of environments; not by sequentially being programmed (or taught) to perform one task in one environment at a time, but rather by intelligently choosing appropriate actions for whatever task and environment it is facing. This goal remains a challenge. In this talk I’ll describe recent work in our lab aimed at the goal of general-purpose robot manipulation by integrating task-and-motion planning with various forms of model learning. In particular, I’ll describe approaches to manipulating objects without prior shape models, to acquiring composable sensorimotor skills, and to exploiting past experience for more efficient planning. Biography: Tomás Lozano-Pérez is Professor in EECS at MIT, and a member of CSAIL. He was a recipient of the 2011 IEEE Robotics Pioneer Award and a co-recipient of the 2021 IEEE Robotics and Automation Technical Field Award. He is a Fellow of the AAAI, ACM, and IEEE.

1 0

Yuanhao Wang will present his General Exam on Friday, October 21, 2022 at 12pm via Zoom.
by Nicki Mahler 20 Oct '22

20 Oct '22

Yuanhao Wang will present his General Exam on Friday, October 21, 2022 at 12pm via Zoom. Zoom link: [ https://princeton.zoom.us/j/4269281435 | https://princeton.zoom.us/j/4269281435 ] The members of his committee are as follows: Chi Jin (adviser), Huacheng Yu, and Jason Lee Abstract Reinforcement learning (RL) is an interactive decision making problem where an agent tries to maximize its cumulative reward through interacting with an unknown environment. RL has seen a surge of interest over the past decade, fueled primarily by its empirical success in a wide range of tasks, including Atari games, Go, robotics and dialogue systems. Motivated by and complementing empirical research, theoretical understanding of RL has also made substantial progress in the past few years. In the most basic case, tabular RL, minimax optimal algorithms have been identified and analyzed. Recent focus of research in the tabular case has shifted toward instance dependent bounds, adversarial settings and multi-agent RL. However, the tabular case fails to capture many real-world problems where the number of states is huge or infinite, in which case function approximation is necessary. In contrast to the usual supervised learning case where complexity measures such as VC dimension are able to characterize the feasibility of learning, the search for structural assumptions that would empower efficient RL remains inconclusive. It has been conjectured and recently proven that realizability alone is insufficient even for linear function approximation, making stronger conditions necessary. Several candidates include low Bellman rank, low witness rank and low Bellman-Eluder dimension, yet the picture is far from complete. My research focus would be on improving the theoretical understanding of RL by addressing some of the aforementioned challenges, such as those about multi-agent learning and function approximation RL. On the multi-agent side, I would like to investigate which solution concepts, such as rationalizability, no-regret and various equilibria, are attainable in RL. On the function approximation side, I hope to identify minimal structural assumptions that permit efficient RL. Reading List: [ https://docs.google.com/document/d/14rchqe0CO9DI9IvmXnFkNX4lf8HahLWtn1tAU0G… | https://docs.google.com/document/d/14rchqe0CO9DI9IvmXnFkNX4lf8HahLWtn1tAU0G… ] Everyone is invited to attend the talk, and those faculty wishing to remain for the oral exam following are welcome to do so.

1 0

Civil & Environmental Engineering speaker: Andrea Castelletti, Tuesday, Nov 1
by Emily C. Lawrence 17 Oct '22

17 Oct '22

Civil & Environmental Engineering Speaker: Andrea Castelletti: Environmental Intelligence Lab, Politecnico di Milano Date: Tuesday, Nov 1, 2022 Time: 4:30PM Location: Louis Simpson, Room A71 Title: Machine Learning advancements for design of water and energy policies in a changing climate and society Abstract: Advances in environmental data monitoring and earth systems’ modeling have increasingly allowed us to accurately reproduce physical processes and their interactions at multiple scales, improving our ability to inform water and energy systems policy design. Yet, many process-based models are limited in predicting complex dynamics, which are the key to strategic planning, such as the impact of extreme weather and climate events or the mutual influence between humans and earth systems. Our work has shown that advanced data analytics and Machine Learning offer new opportunities to better characterize and model coupled human-earth system processes in a world in transition, ultimately supporting more equitable and efficient decision-making processes. In this talk, we provide an overview of recent advances in data-driven modeling and control of human-water-energy systems and showcase how Machine Learning techniques can help (i) infer natural and anthropogenic drivers of observed hydroclimatic patterns and improve their predictability in space and time, (ii) understand and conceptualize the mutual influences between human behaviors and water-energy systems; and (iii) design strategic planning and management policies optimizing multiple and conflicting objectives with different dynamics and informed by heterogeneous information. We recognize that this is a rapidly evolving research field, and we will also stimulate discussion around key challenges, including modeling decisions under uncertainty, model explain ability, data and computational requirements, model scalability, and transferability.

1 0

Princeton Robotics Seminar: Sarah Tang, Fri Oct 14
by Emily C. Lawrence 12 Oct '22

12 Oct '22

Princeton Robotics Seminar Speaker: [ https://www.sarahtang.net/ | Sarah Tang ] , Staff Software Engineer in Planning, Waymo Date: Friday Oct 14, 2022 Time: 11:00am EDT Location: Computer Science 105 Event page: [ https://robo.princeton.edu/seminar | https://robo.princeton.edu/seminar ] Title: Data-Centric ML for Autonomous Driving Abstract: Waymo is building the "World's Most Experienced Driver". This talk will discuss data challenges that arise when scaling machine learning systems with almost 1500 years worth of real-life human driving. Biography: Sarah is passionate about building robots that make intelligent decisions in complex, dynamic environments. She is currently a Staff Software Engineer on the Planner ML team at Waymo, where she is working to make driving safer, smarter, and more scalable with high-capacity models. In previous lives, she was the Tech Lead and Manager of the Motion Planning team at Nuro, where she led development of the decision making and trajectory optimization stack for major product milestones, including autonomous operation of their last-mile delivery robot without a driver nor a chase-car-operator in Arizona, California, and Texas. She was recognized in Business Insider’s 2021 list of “Rising stars in autonomous vehicles”. Sarah got her start in robotics as a member of Princeton's Great Class of 2013.

1 0

Sunayana Rane will present her General Exam "Correspondences between word learning in children and captioning models" on Friday, October 14, 2022 at 1:00 PM on Zoom.
by Louis W. Riehl 11 Oct '22

11 Oct '22

Sunayana Rane will present her General Exam "Correspondences between word learning in children and captioning models" on Friday, October 14, 2022 at 1:00 PM on Zoom. Zoom link: [ https://princeton.zoom.us/my/srane | https://princeton.zoom.us/my/srane ] Committee Members: Tom Griffiths (advisor), Olga Russakovsky, Karthik Narasimhan Abstract: For human children as well as machine learning systems, a key challenge in learning a word is linking the word to the visual phenomena it describes. Today I will present results from a project which aims to understand whether children and machine learning systems have developed similar mechanisms of solving this challenge. By organizing model output into word categories used to analyze child language learning data, we show a correspondence between word learning in children and the performance of image captioning models. Although captioning models are trained only on standard machine learning data, we find that their performance in producing words from a variety of word categories correlates with the age at which children acquire words from each of those categories. To explain why this correspondence exists, we show that the performance of captioning models is correlated with human judgments of the concreteness of words, suggesting that these models are capturing the complex real-world association between words and visual phenomena. Reading List: [ https://docs.google.com/document/d/1NJ4O8NIi2VyTCwKZVLuKgjIl35v8YpSurb_Qhb7… | https://docs.google.com/document/d/1NJ4O8NIi2VyTCwKZVLuKgjIl35v8YpSurb_Qhb7… ] g Everyone is invited to attend the talk, and those faculty wishing to remain for the oral exam following are welcome to do so. Louis Riehl Graduate Administrator Computer Science Department, CS213 Princeton University (609) 258-8014

1 0

Princeton Wireless Distinguished Seminar: Lajos Hanzo, Thursday Oct 13 at 12:30pm
by Emily C. Lawrence 11 Oct '22

11 Oct '22

Princeton Wireless Distinguished Seminar Series Speaker: Lajos Hanzo, University of Southampton, UK Title: Space, Air, Ground Integrated Networking from Single – To Multi-Component Pareto Optimization Day: October 13, 2022 Time: 12:30 pm EST (5:30 pm UK) Location: Zoom (Please Register) [ https://princeton.zoom.us/meeting/register/tJ0rdu6tqz4tEtZH5LmmVelIE-KL5dcw… | https://princeton.zoom.us/meeting/register/tJ0rdu6tqz4tEtZH5LmmVelIE-KL5dcw… ] Hosts: Kyle Jamieson & Yasaman Ghasempour Abstract: Thanks to the spectacular advances in signal processing and nano-technology, five wireless generations have been conceived over the past five decades. Indeed, near-capacity operation at an infinitesimally low error rate has become feasible and flawless multimedia communications is supported in areas of high traffic density, but how do we fill the huge coverage holes existing across the globe? As a promising system architecture, an integrated terrestrial, UAV-aided, airplane-assisted as well as satellite-based global coverage solution will be highlighted to pave the way for seamless next-generation service provision. However, these links exhibit strongly heterogeneous properties, hence requiring different enabling techniques. The joint optimization of the associated conflicting performance metrics of throughput, transmit power, latency, error probability, hand-over probability and link lifetime poses an extremely challenging problem. Explicitly, sophisticated multi-component system optimization is required for finding the Pareto-front of all optimal solutions, where none of the above-mentioned metric can be improved without degrading at least one of the others. Bio: Lajos Hanzo is a Fellow of the Royal Academy of Engineering (FREng), FIEEE, FIET and a EURASIP Fellow, Foreign Member of the Hungarian Academy of Science. He holds honorary Doctorates from the University of Edinburgh and the Technical University of Budapest. He co-authored 19 IEEE Press – John Wiley books and 2000+ research contributions at IEEE Xplore. For further information on his research in progress and associated publications, please refer to IEEE Xplore.

1 0

Haoyu Zhao will present his General Exam, "Coresets for Vertical Federated Learning" on Tuesday, October 11, 2022 at at 1:30 PM in via Zoom.
by Louis W. Riehl 07 Oct '22

07 Oct '22

Haoyu Zhao will present his General Exam, "Coresets for Vertical Federated Learning" on Tuesday, October 11, 2022 at at 1:30 PM in via Zoom. Zoom link: [ https://princeton.zoom.us/j/3247993054 | https://princeton.zoom.us/j/3247993054 ] Committee Members: Sanjeev Arora (advisor), Chi Jin, Elad Hazan Abstract: Vertical federated learning (VFL), where data features are stored in multiple parties distributively, is an important area in machine learning. However, the communication complexity for VFL is typically very high. In this paper, we propose a unified framework by constructing \emph{coresets} in a distributed fashion for communication-efficient VFL. We study two important learning tasks in the VFL setting: regularized linear regression and $k$-means clustering, and apply our coreset framework to both problems. We theoretically show that using coresets can drastically alleviate the communication complexity, while nearly maintain the solution quality. Numerical experiments are conducted to corroborate our theoretical findings. Reading List: [ https://docs.google.com/document/d/1lNANJqOijCA85TNVJ8p95BFrsnVVuP_EOth6T3A… | https://docs.google.com/document/d/1lNANJqOijCA85TNVJ8p95BFrsnVVuP_EOth6T3A… ] Everyone is invited to attend the talk, and those faculty wishing to remain for the oral exam following are welcome to do so. Louis Riehl Graduate Administrator Computer Science Department, CS213 Princeton University (609) 258-8014

1 0

Decentralization of Power Through Blockchain Technology: Happening TODAY
by Emily C. Lawrence 06 Oct '22

06 Oct '22

Join us for the first seminar series event for Princeton's new Center for the [ https://decenter.princeton.edu/ | Decentralization of Power Through Blockchain Technology ] , also known as DeCenter. Lunch will be available beginning at noon. This first event will have two parts: first, the DeCenter co-directors, Dean Andrea Goldsmith and Professor Jaswinder Singh will speak about the goals for the new center. Then, a panel will discuss some of the technology and applications aspirations around blockchains along with their implications for society. Introduction to DeCenter: Andrea J. Goldsmith, Dean of the School of Engineering and Applied Science, Arthur LeGrand Doty Professor of Electrical Engineering, and co-director of DeCenter Jaswinder Singh, Professor of Computer Science, Technology, and Societal Change, co-director of DeCenter Technology, Applications, and Society, a panel discussion moderated by Professor Singh, featuring: Maria Apostolaki, Assistant Professor of Electrical and Computer Engineering Nolan McCarty, Susan Dod Brown Professor of Politics and Public Affairs and Director of the Center for Data-Driven Social Science Andrés Monroy-Hernández, Assistant Professor of Computer Science Pramod Viswanath, Forrest G. Hamrick Professor in Engineering Members of the public (without Princeton affiliation) who wish to attend must register. PUID holders do not need to register. And stay tuned for future DeCenter seminars! Monthly seminars will take place fully in-person on the first Thursday of the month from 12:30 p.m. - 1:20 p.m. in Computer Science Room 105. Videos of seminars will be made available after the event. The next seminar will take place on November 3. [ https://www.cs.princeton.edu/events/26262 | Click here for talk event info. ]

1 0

Arvind Narayanan - The Limits Of The Quantitative Approach To Discrimination: Oct 11
by Emily C. Lawrence 05 Oct '22

05 Oct '22

James Baldwin Lecture Series: Arvind Narayanan – The Limits Of The Quantitative Approach To Discrimination Date: Tuesday, October 11, 2022 Time: 5:00 pm - 6:30 pm Website: [ https://aas.princeton.edu/events/2022/james-baldwin-lecture-series-limits-q… | https://aas.princeton.edu/events/2022/james-baldwin-lecture-series-limits-q… ] Location East Pyne 010 [ https://www.eventbrite.com/e/james-baldwin-lecture-series-tickets-417929748… | REGISTER to attend. ] The annual James Baldwin Lecture series was launched on March 29, 2006, aiming to celebrate the work of Princeton faculty and to provide an occasion for the intellectual community to reflect on the issue of race and American democracy. The lectures also honor the work of the late essayist James Baldwin, one of America’s most powerful cultural critics. Discrimination is obvious to the people facing discrimination. Given this, do we even need quantitative studies to test if it exists? Regardless of the answer, quantitative studies such as ProPublica’s “Machine Bias” have had a galvanizing effect on racial justice, especially in the context of automated decision-making. In this talk, I will discuss what the quantitative approach can reveal, but, more importantly, situations where it cannot tell us what we need to know. Both because of the inherent limits of quantification and because of the way knowledge is socially constructed in quantitative communities, such studies tend to drastically underestimate discrimination, oppression, and algorithmic harms. In the end, quantification is no substitute for centering the experiences of those harmed. Bio: Arvind Narayanan is a professor of computer science at Princeton. His work was among the first to show how machine learning reflects cultural stereotypes including racial and gender biases. He is co-authoring a [ https://fairmlbook.org/ | textbook in fairness and machine learning ] . Narayanan co-created an online course and textbook on bitcoin and cryptocurrency technologies which has been used in over 150 courses worldwide. He is a recipient of the Presidential Early Career Award for Scientists and Engineers (PECASE), twice recipient of the Privacy Enhancing Technologies Award, and thrice recipient of the Privacy Papers for Policy Makers Award.

1 0

Neil Agarwal will present his General Exam on Tuesday, October 4th, 2022 at 1pm in CS 105.
by Nicki Mahler 30 Sep '22

30 Sep '22

Neil Agarwal will present his General Exam on Tuesday, October 4th, 2022 at 1pm in CS 105. Zoom link: https://princeton.zoom.us/j/95202734109 The members of his committee are as follows: Ravi Netravali (adviser), Jennifer Rexford, and Amit Levy Abstract: https://docs.google.com/document/d/1QRnVxlcKWewGSNBl19IARoH4gCg2YvISSeqJ1oj… Machine learning-based video analytics pipelines have become the de facto solution for automated and accurate video processing. However, deploying video analytics pipelines has become increasingly prohibitive due to the growing number of camera feeds, more complex and specialized vision models, and the broadening range of applications and use cases. Practically supporting this rising demand necessitates the efficient utilization of underlying resources (e.g., heterogeneous hardware, storage, network bandwidth). My research focuses on designing automated and general resource optimization techniques and building systems to practically incorporate these techniques into existing video analytics pipelines. My work has targeted 3 key operational scenarios: querying large-scale video datasets in the cloud, real-time analytics in resource-constrained edge settings, and privacy-preserving video analytics queries. In today's talk, I will discuss one of these projects, Gemel, which aims to reduce GPU memory pressure in edge video analytics deployments. We observe that edge-box GPUs lack the memory needed to concurrently house the growing number of (increasingly complex) models for real-time inference. We present model merging, a new memory management technique that exploits architectural similarities between edge vision models by judiciously sharing their layers (including weights) to reduce workload memory costs and swapping delays. Gemel efficiently integrates merging into existing pipelines by (1) leveraging several guiding observations about per-model memory usage and inter-layer dependencies to quickly identify fruitful and accuracy-preserving merging configurations, and (2) altering edge inference schedules to maximize merging benefits. Reading List: https://docs.google.com/document/d/1QRnVxlcKWewGSNBl19IARoH4gCg2YvISSeqJ1oj…

1 1