February 2022 - talks - lists.cs.princeton.edu

DataX Workshop Series: Bridging the Gap Between Your Laptop and Cloud Computing
by Emily C. Lawrence 28 Feb '22

28 Feb '22

DataX Workshop Series: Bridging the Gap Between Your Laptop and Cloud Computing [ https://csml.princeton.edu/people/jose-garrido-torres | Jose Garrido Torres ] , DataX Catalysis Data Scientist [ https://csml.princeton.edu/news/solving-technical-debt-dilemma-princetons-m… | Vineet Bansal ] , Sr. Research Software Engineer March 4, 2022 April 1, 2022 9:00 – 10:00 a.m. Part 1: Introduction to some tools that computer programmers typically use to write and debug code in an effective manner. Part 2: Introduction to cloud computing (create and manage cloud computing resources) How to use some tools that offer the possibility of writing code locally while seamlessly executing/running it on powerful cloud computing. Location : 26 Prospect Avenue, Room 103 In Person Attendance is limited and for Princeton University affiliates . This presentation will also be available via Zoom. Please note if you are attending virtually and a link will be provided. Part 1: MARCH 4 Workshop: [ https://docs.google.com/forms/d/e/1FAIpQLSfYeMMKdkUWHiuer0Aw7uZPwcvnaYOSPJO… | Register Here ] . Part 2: APRIL 1 Workshop: [ https://docs.google.com/forms/d/e/1FAIpQLSdDGuhgSKLt5HEztbNd1GFm-leAcxMUXzG… | Register Here ] .

1 0

Sobhan Miryoosefi will present his Pre FPO "Provable Reinforcement Learning with Constraints and Function Approximation" on Friday, March 4, 2022 at 1pm via Zoom.
by Nicki Mahler 22 Feb '22

22 Feb '22

Sobhan Miryoosefi will present his Pre FPO "Provable Reinforcement Learning with Constraints and Function Approximation" on Friday, March 4, 2022 at 1pm via Zoom. Zoom Link: https://princeton.zoom.us/j/5269962765 Committee Members: Examiners: Chi Jin (ECE, advisor), Matthew Weinberg, Karthik Narasimhan Readers: Elad Hazan, Jason Lee (ECE) All are welcome to attend. Abstract: Reinforcement learning (RL) has gained a surge of interest over the past years, fueled mainly by practical success and new applications in various domains. However, there is still a gap between our theoretical understanding of these RL techniques and their empirical success. In this thesis, we advance our understanding by studying reinforcement learning from a primarily theoretical point of view and designing provably efficient algorithms for two challenging settings of 1) reinforcement learning with constraints and 2) reinforcement learning with function approximation. 1) In standard reinforcement learning (RL), a learning agent seeks to optimize the overall reward. However, many key aspects of the desired behavior are more naturally expressed as constraints. For instance, the designer may want to limit the use of unsafe actions, increase the diversity of trajectories to enable exploration, or approximate expert trajectories when rewards are sparse. First, we propose an algorithmic scheme that can handle RL tasks with general convex constraints improving upon prior works that are either limited to linear constraints or lack theoretical guarantees. Second, focusing on sample-efficient exploration, we develop the first provably efficient algorithm for tabular episodic constrained RL with the ability to handle convex constraints as well as the knapsack setting. Finally, motivated by recent advances in reward-free RL, we propose a simple meta-algorithm such that given any reward-free RL oracle, the constrained RL problems can be directly solved with negligible overheads in sample complexity. 2) Finding the minimal structural assumptions that empower sample-efficient learning is one of RL's most important research directions. This thesis advances our understanding of this fundamental question by introducing a new complexity measure---Bellman Eluder (BE) dimension. We show that the family of RL problems of low BE dimension is remarkably rich, which subsumes a vast majority of existing tractable RL problems, including but not limited to tabular MDPs, linear MDPs, reactive POMDPs, low Bellman rank problems as well as low Eluder dimension problems. We further design a new optimization-based algorithm—GOLF, and reanalyze a hypothesis elimination-based algorithm—OLIVE (proposed in Jiang et al., 2017) . We prove that both algorithms learn the near-optimal policies of low BE dimension problems in a number of samples that is polynomial in all relevant parameters but independent of the size of state-action space. Our regret and sample complexity results match or improve the best existing results for several well-known subclasses of low BE dimension problems. Furthermore, moving towards a more challenging setting of partial observable RL, we study a new subclass of Partially Observable Markov Decision Processes (POMDPs). More specifically, motivated by the problem structure in several physical applications and a commonly used technique known as "frame stacking'', this work proposes to study a new subclass of POMDPs, whose latent states can be decoded by the most recent history of a short length m. We establish a set of upper and lower bounds on the sample complexity for learning near-optimal policies for this class of problems in both tabular and rich-observation settings (where the number of observations is enormous). Our results show that a short-term memory suffices for reinforcement learning in these environments.

1 0

Keith Winstein: What we talk about when we talk about networking. Tuesday, February 22nd, 12pm in CS 105
by Emily C. Lawrence 18 Feb '22

18 Feb '22

Guest Speaker Speaker: Keith Winstein, Stanford Date: Tuesday, Feb 22, 2022 Time: 12:00pm EST Host: Prof. Amit Levy Location: CS 105 Title: What we talk about when we talk about networking Abstract: Networks, and the applications they support, sometimes treat each other as strangers. By shaking things up a bit—expressing networked systems as compositions of small, pure functions and making their dataflow a first-class consideration—we can often achieve friendlier couplings across the stack, to the benefit of performance, robustness, and understandability. This approach has proved helpful in several contexts: networking algorithms learned "in situ," feeding data from deployment back into training; real-time video conferencing, especially for musicians and actors during the pandemic; image compression in a distributed network filesystem; and a serverless computing framework that lets software burst briefly to 10,000 cores. In ongoing work, we're building a "functional" operating system that enforces a separation between IO (declared to the OS) and computation (reproducible by default). If this system can support a broad range of computational tasks with visibility into their dataflow, we envision a new service model for cloud computing: "computation as a service." Bio: Keith Winstein is an assistant professor of computer science and, by courtesy, of electrical engineering at Stanford University. His research group creates new kinds of networked systems by rethinking abstractions around communication, compression, and computing. Some of his group’s research has found broader use, including the Mosh tool, the Puffer video-streaming site, the Lepton compression tool, and the Mahimahi network emulators. Winstein previously served as a staff reporter at The Wall Street Journal and worked at Ksplice, a startup company (now part of Oracle) where he was the vice president of product management and business development and also cleaned the bathroom. He did his undergraduate and graduate work at MIT. (https://cs.stanford.edu/~keithw)

1 0

Colloquium Speaker Yael Tauman Kalai: UPDATE - Livestream unavailable
by Emily C. Lawrence 17 Feb '22

17 Feb '22

My apologies. We have technical problems in the lecture hall so the talk cannot be live-streamed. However a recording will be made available on the CS Youtube page here: [ https://www.youtube.com/channel/UC5tTXQLklM6OQlN1kl7cKXQ | https://www.youtube.com/channel/UC5tTXQLklM6OQlN1kl7cKXQ ] Regards, Emily From: "Emily Lawrence" <emilyl(a)cs.princeton.edu> To: "talks" <talks(a)lists.cs.princeton.edu> Sent: Thursday, February 17, 2022 9:00:00 AM Subject: [talks] Colloquium Speaker Yael Tauman Kalai: TODAY at 12:30pm CS Colloquium Speaker Speaker: Yael Tauman Kalai, Microsoft Research and MIT Date: Thursday, Feb 17, 2022 Time: 12:30pm EST Host: Prof. Ran Raz Location: Live-stream online Event page: [ https://www.cs.princeton.edu/events/26158 | https://www.cs.princeton.edu/events/26158 ] View live-stream online here: [ https://mediacentrallive.princeton.edu/ | https://mediacentrallive.princeton.edu/ ] Title: Efficient Verification of Computation Abstract: Efficient verification of computation is fundamental to computer science, and is at the heart of the P vs. NP question. Recently it has had growing practical significance, especially with the increasing popularity of blockchain technologies and cloud computing. In this talk, I will present schemes for verifying the correctness of a computation. I will discuss both their practical aspects and their impact on quantum complexity, hardness of approximation, and the complexity of Nash equilibrium. Bio: Yael Tauman Kalai received her BA (1997) from the Hebrew University in Jerusalem, MA (2001) under the supervision of Adi Shamir at the Weizmann Institute, and PhD (2006) under the supervision of Shafi Goldwasser at MIT. After postdoctoral positions at Microsoft Research and the Weizmann Institute, she is now a Principal Senior Researcher at Microsoft Research New England and an adjunct professor at MIT. Her research focuses on cryptography. _______________________________________________ talks mailing list talks(a)lists.cs.princeton.edu To edit subscription settings or remove yourself, use this link: https://lists.cs.princeton.edu/mailman/listinfo/talks

1 0

Colloquium Speaker Yael Tauman Kalai: Thursday, Feb 17 at 12:30pm
by Emily C. Lawrence 17 Feb '22

17 Feb '22

CS Colloquium Speaker Speaker: Yael Tauman Kalai, Microsoft Research and MIT Date: Thursday, Feb 17, 2022 Time: 12:30pm EST Host: Prof. Ran Raz Location: Live-stream online Event page: [ https://www.cs.princeton.edu/events/26158 | https://www.cs.princeton.edu/events/26158 ] View live-stream online here: [ https://mediacentrallive.princeton.edu/ | https://mediacentrallive.princeton.edu/ ] Title: Efficient Verification of Computation Abstract: Efficient verification of computation is fundamental to computer science, and is at the heart of the P vs. NP question. Recently it has had growing practical significance, especially with the increasing popularity of blockchain technologies and cloud computing. In this talk, I will present schemes for verifying the correctness of a computation. I will discuss both their practical aspects and their impact on quantum complexity, hardness of approximation, and the complexity of Nash equilibrium. Bio: Yael Tauman Kalai received her BA (1997) from the Hebrew University in Jerusalem, MA (2001) under the supervision of Adi Shamir at the Weizmann Institute, and PhD (2006) under the supervision of Shafi Goldwasser at MIT. After postdoctoral positions at Microsoft Research and the Weizmann Institute, she is now a Principal Senior Researcher at Microsoft Research New England and an adjunct professor at MIT. Her research focuses on cryptography.

1 2

Kaiyu Yang will present his Pre FPO on Tuesday, February 22, 2022 at 10am in Friend Center 125 and Zoom (hybrid)
by Nicki Mahler 16 Feb '22

16 Feb '22

Kaiyu Yang will present his Pre FPO on Tuesday, February 22, 2022 at 10am in Friend Center 125 and Zoom (hybrid) Zoom link: https://princeton.zoom.us/j/2731344683 Committee members: Jia Deng (examiner, adviser), Olga Russakovsky (examiner), Danqi Chen (examiner), Karthik Narasimhan (reader), Mayur Naik (reader, University of Pennsylvania) Abstract: Symbolic reasoning is an important tool for modeling our world. For example, physics is essentially constructing symbolic models of the natural world, and engineering focuses on using such models to manipulate the world. Current machine learning is not capable of constructing such compact and powerful models, and is not capable of symbolic reasoning. Advances in symbolic reasoning can be revolutionary, potentially leading to machines that discover new math, sciences, and algorithms. My long-term goal is to make machine learning capable of symbolic reasoning. And I try to combine the classical, symbolic approach to AI with the modern, machine learning–based approach. In this talk, I will present my work approaching this problem from two angles: (1) applying machine learning to symbolic reasoning tasks, such as automated theorem proving; (2) introducing symbolic components into machine learning models to make them more interpretable, verifiable, and data-efficient. The core idea behind my solutions is representing symbolic reasoning as proofs, which provides effective inductive biases for reasoning.

1 0

CDH event: Machine Learning & Cultural Heritage, Feb 22 at 4:30pm
by Emily C. Lawrence 16 Feb '22

16 Feb '22

[ https://princeton.us16.list-manage.com/vcard?u=a77a39541c9fc9c8b625cb379&id… ] Novel Machine Learning Methods for Computing Cultural Heritage: An Interdisciplinary Approach We hope you will join us on Tuesday, February 22 at 4:30 pm for an event with [ https://princeton.us16.list-manage.com/track/click?u=a77a39541c9fc9c8b625cb… | Benjamin Lee ] (University of Washington). Lee will share how his project, [ https://princeton.us16.list-manage.com/track/click?u=a77a39541c9fc9c8b625cb… | Newspaper Navigator ] , re-imagines how humanists, social scientists, and the public can navigate and analyze the visual content in millions of digitized historic newspaper pages. He will also introduce his ongoing work surrounding the development of open faceted search systems for petabyte-scale web archives and elaborate on how his research can extend to a wide range of digitized and born-digital collections. ( [ https://princeton.us16.list-manage.com/track/click?u=a77a39541c9fc9c8b625cb… | learn more ] and [ https://princeton.us16.list-manage.com/track/click?u=a77a39541c9fc9c8b625cb… | register ] ) The event will be moderated by [ https://princeton.us16.list-manage.com/track/click?u=a77a39541c9fc9c8b625cb… | Jim Casey ] (Penn State) and [ https://princeton.us16.list-manage.com/track/click?u=a77a39541c9fc9c8b625cb… | Tianyi Wang ] (Princeton). This event is part of the [ https://princeton.us16.list-manage.com/track/click?u=a77a39541c9fc9c8b625cb… | Machine Learning + Humanities Working Group ] Series and is co-sponsored by the [ https://princeton.us16.list-manage.com/track/click?u=a77a39541c9fc9c8b625cb… | Data-Driven Social Science Initiative ] and the [ https://princeton.us16.list-manage.com/track/click?u=a77a39541c9fc9c8b625cb… | Center for Statistics and Machine Learning ] . You are also invited to join the ML + Humanities Working Group for a discussion of the event on February 23 at 12:30 pm ( [ https://princeton.us16.list-manage.com/track/click?u=a77a39541c9fc9c8b625cb… | details on our website ] ).

1 0

Kevin Lee will present his Pre-FPO "Discovering Real-World Weaknesses in User Authentication" on February 25, 2022 at 2pm via Zoom.
by Nicki Mahler 15 Feb '22

15 Feb '22

Kevin Lee will present his Pre-FPO "Discovering Real-World Weaknesses in User Authentication" on February 25, 2022 at 2pm via Zoom. Zoom link: [ https://princeton.zoom.us/j/93407271043 | https://princeton.zoom.us/j/93407271043 ] Committee members: Arvind Narayanan (adviser), Examiners: Andrés Monroy-Hernández, and Tom Ristenpart (Cornell); Readers: Jonathan Mayer and Prateek Mittal. Abstract: The theory and practice of user authentication has come a long way in the last decade. Yet these gains have been uneven. From SIM swaps to romance scams, people continue to fall prey to low-tech but devastating security failures. In this talk, I’ll discuss three studies which uncovered fundamental weaknesses in widely-used user authentication methods. SIM swap attacks allow criminals to hijack a victim’s phone number, which are used to receive one-time passcodes for 2FA. To protect against hijacking, mobile carriers should have procedures in place to properly authenticate a customer calling in to request a SIM swap. Instead, we found that five carriers in the United States continue to use authentication methods that are now known to be insecure, enabling straightforward SIM swap attacks. Through persistent outreach with policymakers and stakeholders, we have been able to effect ongoing rulemaking that aims to protect users against SIM swap attacks. Phone number recycling presents another challenge to using SMS-based authentication. We presented and empirically evaluated attacks enabled by number recycling, a standard industry practice in the U.S. We simulated a UI-bound adversary and looked for vulnerable available recycled numbers at mobile carriers. We found most numbers were vulnerable, which could subject previous owners to account hijackings and privacy risks. While carriers, websites, and subscribers can take steps to reduce risk, number recycling threats highlight fundamental problems with the use of phone numbers for security-sensitive purposes. Finally, I’ll discuss recent work looking at password policies. We examined the policies of 120 of the most popular websites for when a user creates a new password for their account. Despite well-established advice that has emerged from the research community, we found that only 11% of websites followed all relevant best practices in their password policies. Specifically, 75% of websites do not stop users from choosing the most common passwords, while 45% burden users by requiring specific character classes in their passwords for minimal security benefit. Worse, we found low adoption of password strength meters---a widely touted intervention to encourage stronger passwords, appearing on only 19% of websites. I’ll discuss the implications of our discoveries, and propose avenues for future research to understand the disconnect between password research and industry practices.

1 0

Zhiyuan Li will present his Pre FPO "Toward Mathematical Understanding of Real-life Deep Learning" on February 18, 2022 at 2:30pm via Zoom
by Nicki Mahler 14 Feb '22

14 Feb '22

Zhiyuan Li will present his Pre FPO "Toward Mathematical Understanding of Real-life Deep Learning" on February 18, 2022 at 2:30pm via Zoom Zoom link: https://princeton.zoom.us/j/93110423178?pwd=K2RoRm9YTmhacE5pU3JnZE1JbmdQdz09 Committee: Sanjeev Arora (adviser), Danqi Chen (non-reader), Chi Jin(non-reader, ECE), Elad Hazan (Reader), Jason Lee( ECE, Reader). Title: Toward Mathematical Understanding of Real-life Deep Learning Abstract: There is great interest in developing a mathematical understanding of the tremendous success of deep learning. Most of this understanding has been done in simplified settings (depth 2 or 3; NTK regime). This talk presents my recent works providing a mathematical understanding of real-life nets and losses, incorporating the effect of normalization, architectural features, stochasticity, and finite learning rate(LR). It leverages insights from continuous mathematics (including Stochastic Differential Equation(SDE)) which I will use to show interesting new mechanisms for implicit regularization during training. I will finish by presenting a new practical advance from our theoretical insights: a robust variant of BERT (a language model at the heart of the ongoing revolution in Natural Language Processing) called SIBERT that uses a new scale-invariance architecture and is trainable with vanilla SGD.

1 0

Public lecture: the religious implications of AI with Dr. Will Penman, Wed 2/16 at 7pm
by Emily C. Lawrence 14 Feb '22

14 Feb '22

Who Are You When You Lose? Game-Playing AI, Humility, and Faith With Will Penman, Lecturer, Princeton Writing Program Wednesday, Feb 16 at 7:00pm Frist Campus Center, room 302 Humans aren’t the fastest creatures. We can’t fly, or make food from sunlight, or live at the bottom of the ocean. But when it comes to language, creativity, rationality – those let us overcome our other limitations, and we’re just the best… right? In a few cases, today’s AI is actually beating us already, and it may get better at other cases soon. Examples include text-production AI that can write coherent paragraphs, robots that can backflip, and game-playing AI that beats the world’s best chess players. What do we make of these developments as Christians? How can we tie in our faith when emerging AI sometimes threatens the way we’ve typically seen ourselves? First, we’ll offer some wider historical context where scientific discoveries have displaced how central humans are in the universe. Then, we’ll examine a Christian life of humility as a key to responding well to AI displacement. Humility gives us a flexibility of self that allows us to hear God’s countercultural call. In the case where we’re losing to AI at games we care about, being humble means internalizing that humans are not necessarily the height of intelligence/rationality/etc. We apply this tough message by considering situations in daily life where identifying too closely with our own intelligence might be holding us back, and where God is calling us into deeper humility.

1 0