April 2023 - talks - lists.cs.princeton.edu

Daniel Suo will present his FPO "Scaling Machine Learning in Practice" on Wednesday, May 10, 2023 at 3pm in CS 402.
by Nicki Mahler 10 May '23

10 May '23

Daniel Suo will present his FPO "Scaling Machine Learning in Practice" on Wednesday, May 10, 2023 at 3pm in CS 402. The members of his committee are as follows: Examiners: Kai Li (Adviser), Olga Troyanskaya, and Ryan Adams; Readers: Elad Hazan (Co-Adviser) and Naman Agarwal (Google LLC) All are welcome to attend. In recent years, machine learning has become pervasive, powering algorithmic clinicians, translators, and world-beating go masters. As practitioners build on this success, they repeatedly observe that scale–data, model size, compute–is critical. However, scale is now a challenge in and of itself; simple tasks such as gathering data become formidable, even prohibitive. In this talk, we discuss techniques for addressing scale in three areas: 1. Differential reinforcement learning for physical devices: reinforcement learning has emerged as a potential strategy for machines to make decisions in complex, dynamic environments. However, successful demonstrations have required vast experience to learn an optimal policy, making real-world physical applications particularly challenging. We present a method that uses limited experience to learn a differentiable simulator of a physical system (medical ventilator) and then uses gradient methods on the simulator to learn a state-of-the-art policy for controlling that system. 2. Practical optimization for deep learning: optimization is an essential aspect of deep learning. However, while a constellation of optimization algorithms dot the literature, the low burden of proof and empirical nature of deep learning has led practitioners to rely on defaults (i.e., Adagrad, Adam) rather than view optimization as a lever for progress. To rigorously test ideas in optimization, we introduce a comprehensive benchmark that currently includes 8 deep learning workloads and rules for training procedures, computational budget, and evaluation. 3. Scaling computer systems via thread scheduling: large global-scale applications are expensive and complex to operate let alone optimize. As a result, many simple parameters that govern important behaviors of these systems are simply set once and never touched again. However, we show that these parameters present low-hanging fruit for significant eciency improvements

1 1

Art of Science 2023 Opening Reception: Tuesday, May 2 at 4:30pm
by Emily C. Lawrence 02 May '23

02 May '23

[ https://artofsci.princeton.edu/ | Art of Science 2023 Opening Reception ] Join us Tuesday to celebrate the opening of this year’s Art of Science exhibition at the Friend Center. Winners will be announced, food and drink will be served. Tuesday, May 2, 2023 4:30-6:00pm Main Atrium, Friend Center ( [ https://goo.gl/maps/pYaixxBipNBbD74h6 | Map It ] ) Princeton University Celebrating its 17th year, Art of Science explores the interplay between science and art. This year’s exhibition features still images, videos and animations from 50 creators—including students, alumni, staff and faculty from across the Princeton University community. Visit [ https://artofsci.princeton.edu/ | artofsci.princeton.edu ] for details. We look forward to seeing you at the event! The Exhibition Organizers

1 1

CS Colloquium Speaker Lydia Liu TODAY at 12:30pm
by Emily C. Lawrence 01 May '23

01 May '23

Speaker: Lydia Liu, Cornell University Date: Wednesday, April 5 Time: 12:30pm EST Location: CS 105 Hosts: Olga Russakovsky and Peter Ramadge Event page: https://www.cs.princeton.edu/events/26353 Title: Towards Responsible Machine Learning in Societal Systems Abstract: Machine learning systems are deployed in consequential domains such as education, employment, and credit, where decisions have profound effects on socioeconomic opportunity and life outcomes. High stakes decision settings present new statistical, algorithmic, and ethical challenges. In this talk, we examine the distributive impact of machine learning algorithms in societal contexts, and investigate the algorithmic and sociotechnical interventions that bring machine learning systems into alignment with societal values---equity and long-term welfare. First, we study the dynamic interactions between machine learning algorithms and populations, for the purpose of mitigating disparate impact in applications such as algorithmic lending and hiring. Next, we consider data-driven decision systems in competitive environments such as markets, and devise learning algorithms to ensure efficiency and allocative fairness. We end by outlining future directions for responsible machine learning in societal systems that bridge the gap between the optimization of predictive models and the evaluation of downstream decisions and impact. Bio: Lydia T. Liu is a postdoctoral researcher in Computer Science at Cornell University, working with Jon Kleinberg, Karen Levy, and Solon Barocas. Her research examines the theoretical foundations of machine learning and algorithmic decision-making, with a focus on societal impact and human welfare. She obtained her PhD in Electrical Engineering and Computer Sciences from UC Berkeley, advised by Moritz Hardt and Michael Jordan, and has received a Microsoft Ada Lovelace Fellowship, an Open Philanthropy AI Fellowship, an NUS Development Grant, and a Best Paper Award at the International Conference on Machine Learning. This talk is co-sponsored with Electrical and Computer Engineering and the Center for Information Technology Policy.

1 8

Mike Wong will present his General Exam "MadEye: Boosting Live Video Analytics Accuracy with Adaptive Camera Configurations" on Monday, May 15, 2023 at 1pm in CS 302.
by Nicki Mahler 28 Apr '23

28 Apr '23

Mike Wong will present his General Exam "MadEye: Boosting Live Video Analytics Accuracy with Adaptive Camera Configurations" on Monday, May 15, 2023 at 1pm in CS 302. The members of his committee are as follows: Ravi Netravali (adviser), Amit Levy, and Wyatt Lloyd Abstract: Camera orientations (i.e., rotation and zoom) govern the content that a camera captures in a given scene, which in turn heavily influences the accuracy of live video analytics pipelines. However, existing analytics approaches leave this crucial adaptation knob untouched, instead opting to only alter the way that captured images from fixed orientations are encoded, streamed, and analyzed. We present MadEye, a camera-server system that automatically and continually adapts orientations to maximize accuracy for the workload and resource constraints at hand. To realize this using commodity pan-tilt-zoom (PTZ) cameras, MadEye embeds (1) a search algorithm that rapidly explores the massive space of orientations to identify a fruitful subset at each time, and (2) a novel knowledge distillation strategy to efficiently (with only camera resources) select the ones that maximize workload accuracy. Experiments on diverse workloads show that MadEye boosts accuracy by 2.9-25.7% for the same resource usage, or achieves the same accuracy with 2-3.7× lower resource costs. Link to abstract and reading list can be found here: [ https://docs.google.com/document/d/1bQBds_E_Iiq0D8pIM0MYTV4gakVTCBWa1bUh4my… | https://docs.google.com/document/d/1bQBds_E_Iiq0D8pIM0MYTV4gakVTCBWa1bUh4my… ] 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

CITP Special Event: Thursday, May 4 at 4:30pm
by Emily C. Lawrence 28 Apr '23

28 Apr '23

CITP Special Event: Confused by All the Chatter? Journalists, Researchers & Policymakers Talk Chatbots and Other Large Language Models Date: Thursday, May 4, 2023 Time: 4:30 pm - 6:00 pm Location: Lewis Auditorium, Robertson Hall Princeton, 08540 United States [ https://www.google.com/maps/search/?api=1&query=40.3815302%2C-74.651754 | + Google Map ] As part of the CITP Digital Investigators Conference the public is invited to join us in person or virtually for this event. Please register [ https://docs.google.com/forms/d/1VgdnePbpgNWi-xdBNMdXxSrwS3pdiDxtz_twsv_jgV… | here ] to attend in person. The livestream will be available [ https://mediacentrallive.princeton.edu/ | here. ] Powerful new technologies like [ https://openai.com/blog/chatgpt | OpenAI’s “ChatGPT” ] or [ https://blog.google/technology/ai/bard-google-ai-search-updates/ | Google’s “Bard” ] have sparked excitement over the potential they have to transform how we work, learn and communicate for the better. But their potential harms also trigger fears and unease. As a result, the public discourse around such large language models (LLMs) can be noisy or chaotic. CITP has convened a panel of experts from the journalism, tech research and public policy sectors to discuss their experiences with – and approaches to – engaging with these emerging technologies in their respective professions. We will also talk about the responsibilities journalists and academics may have in shaping the public conversation around digital technologies, and how they can support each other’s work for the benefit of the public. Panel Speakers: [ https://juliaangwin.com/ | Julia Angwin ] is an award-winning investigative journalist and contributing opinion writer at The New York Times. She founded The Markup, a nonprofit newsroom that investigates the impacts of technology on society, and is Entrepreneur in Residence at Columbia Journalism School’s Brown Institute. Angwin was previously a senior reporter at the independent news organization ProPublica, where she led an investigative team that was a finalist for a Pulitzer Prize in Explanatory Reporting in 2017 and won a Gerald Loeb Award in 2018. From 2000 to 2013, she was a reporter at The Wall Street Journal, where she led a privacy investigative team that was a finalist for a Pulitzer Prize in Explanatory Reporting in 2011 and won a Gerald Loeb Award in 2010. In 2003, she was on a team of reporters at The Wall Street Journal that was awarded the Pulitzer Prize in Explanatory Reporting for coverage of corporate corruption. She is also the author of “Dragnet Nation: A Quest for Privacy, Security and Freedom in a World of Relentless Surveillance” (Times Books, 2014) and “Stealing MySpace: The Battle to Control the Most Popular Website in America” (Random House, March 2009). She earned a B.A. in mathematics from the University of Chicago, and an M.B.A. from the Graduate School of Business at Columbia University. [ http://sorelle.friedler.net/ | Sorelle Friedler ] is the Shibulal Family Associate Professor of Computer Science at Haverford College. She served as the assistant director for Data and Democracy in the White House Office of Science and Technology Policy under the Biden-Harris Administration where her work included the Blueprint for an AI Bill of Rights. Her research focuses on the fairness and interpretability of machine learning algorithms, with applications from criminal justice to materials discovery. She holds a Ph.D. in computer science from the University of Maryland, College Park. [ https://www.cs.princeton.edu/~arvindn/ | Arvind Narayanan ] is a professor of computer science at Princeton University. He co-authored a textbook on fairness and machine learning and is currently co-authoring a book on AI snake oil. He led the Princeton Web Transparency and Accountability Project to uncover how companies collect and use our personal information. His work was among the first to show how machine learning reflects cultural stereotypes, and his doctoral research showed the fundamental limits of de-identification. Narayanan is a recipient of the Presidential Early Career Award for Scientists and Engineers (PECASE), twice a recipient of the Privacy Enhancing Technologies Award, and thrice a recipient of the Privacy Papers for Policy Makers Award. To request accommodations for a disability please contact Jean Butcher, butcher(a)princeton.edu , at least one week prior to the event. This seminar will be recorded and posted to the CITP website, YouTube channel and Princeton University’s Media Central channel.

1 0

Alex Guerra will present his General Exam "Neuro-Symbolic Function Representations" on Wednesday, May 3, 2023 at 9:00 AM in CS 402 and via zoom.
by Louis W. Riehl 28 Apr '23

28 Apr '23

Alex Guerra will present his General Exam "Neuro-Symbolic Function Representations" on Wednesday, May 3, 2023 at 9:00 AM in CS 402 and via zoom. Zoom link: https://princeton.zoom.us/j/9379621328 Committee Members: Ryan Adams (advisor), Jia Deng, Tom Griffiths Abstract: Science is driven by the quest to find patterns in the universe. Patterns can be expressed as functions, with inputs that give rise to some output. In machine learning, we typically aim to find useful representations of data. However, because the universe is most naturally described by the functions giving rise to this data, we argue that being able to represent functions themselves is a more general, and hence powerful capability. By creating representations of functions themselves, we can learn operators on top of these functions that give us a powerful set of tools to solve problems in new ways. In this work, we introduce a novel method of representing functions, along with a way of learning operators on top of these representations such as composition, differentiation, and integration. We demonstrate how these tools can help us solve a wide array of scientific problems in new ways. The results shown here are merely a small sample of the capabilities of these representations can afford us, and we hope that this work serves to inspire the reader to think about their own problems from a new perspective. At its core, this method is a highly general way of modelling interaction, and we hope that it inspires a new way of thinking about representing symbolic structures. Reading List: https://docs.google.com/document/d/1Y67d2G0fRF-jxLLYB3pop0LQV9V285LE-ErNl6B… 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

Watson Jia will present his MSE talk "QUICstep: Circumventing QUIC-based Censorship" on Friday, April 28, 2023 at3pm in Sherrerd Hall 306.
by Nicki Mahler 26 Apr '23

26 Apr '23

Watson Jia will present his MSE talk "QUICstep: Circumventing QUIC-based Censorship" on Friday, April 28, 2023 at3pm in Sherrerd Hall 306. The members of his committee as as follows: Prateek Mittal (adviser) and Jonathan Mayer (reader) All are welcome to attend. Talk abstract follows below. Abstract: Governments around the world limit free and open communication on the Internet through censorship. To reliably identify and block access to certain web domains, censors inspect the plaintext TLS SNI field sent in TLS handshakes. With QUIC rapidly displacing TCP as the dominant transport-layer protocol on the web, censorship regimes have already begun prosecuting network traffic delivered over QUIC. With QUIC censorship poised to expand, censorship circumvention tools must similarly adapt. We present QUICstep, a censorship-resilient, application-agnostic, performant, and easy-to-implement approach to censorship circumvention in the QUIC era. QUICstep circumvents TLS SNI censorship by conducting a QUIC-TLS handshake over an encrypted tunnel to hide the SNI field from censors and performs connection migration to resume the QUIC session in plain sight of the censor. Our evaluation finds that QUICstep successfully establishes QUIC sessions in the presence of a proof-of-concept censor with minimal latency overhead.

1 0

Yebin Chon will present her General Exam "A Unified API for Memory Dependence Analysis Tools" on Monday, May 1, 2023 at 10am in Friend 202 (CORRECTION)
by Nicki Mahler 26 Apr '23

26 Apr '23

Yebin Chon will present her General Exam "A Unified API for Memory Dependence Analysis Tools" on Monday, May 1, 2023 at 10am in Friend 202. The members of Yebin's committee is as follows: David August (adviser), Amit Levy, and Zachary Kincaid Reading list and abstracts can be found via the links below. Abstract: [ https://docs.google.com/document/d/1Sd5IqojKpFB24PLRgmyqdrxPRrmodts-0TF_w2K… | https://docs.google.com/document/d/1Sd5IqojKpFB24PLRgmyqdrxPRrmodts-0TF_w2K… ] Reading List: [ https://docs.google.com/document/d/1m0vGIihX2A60Z1xRnkTnH7XxSWf7Wn669k9A5tb… | https://docs.google.com/document/d/1m0vGIihX2A60Z1xRnkTnH7XxSWf7Wn669k9A5tb… ] 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

Yebin Chon will present his General Exam "A Unified API for Memory Dependence Analysis Tools" on Monday, May 1, 2023 at 10am in Friend 202.
by Nicki Mahler 26 Apr '23

26 Apr '23

Yebin Chon will present his General Exam "A Unified API for Memory Dependence Analysis Tools" on Monday, May 1, 2023 at 10am in Friend 202. The members of Yebin's committee is as follows: David August (adviser), Amit Levy, and Zachary Kincaid Reading list and abstracts can be found via the links below. Abstract: [ https://docs.google.com/document/d/1Sd5IqojKpFB24PLRgmyqdrxPRrmodts-0TF_w2K… | https://docs.google.com/document/d/1Sd5IqojKpFB24PLRgmyqdrxPRrmodts-0TF_w2K… ] Reading List: [ https://docs.google.com/document/d/1m0vGIihX2A60Z1xRnkTnH7XxSWf7Wn669k9A5tb… | https://docs.google.com/document/d/1m0vGIihX2A60Z1xRnkTnH7XxSWf7Wn669k9A5tb… ] 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

Austin Wang will present his General Exam "Semantic Supervision for Simple and Scalable Zero-shot Generalization" on Monday, May 1, 2023 at 1:30 PM in CS 401 .
by Louis W. Riehl 26 Apr '23

26 Apr '23

Austin Wang will present his General Exam "Semantic Supervision for Simple and Scalable Zero-shot Generalization" on Monday, May 1, 2023 at 1:30 PM in CS 401 . Committee Members: Karthik Narasimhan(advisor), Olga Russakovsky, Danqi Chen Abstract: Zero-shot learning is the problem of predicting instances over classes not seen during training. One approach to zero-shot learning is providing auxiliary class information to the model. Prior work along this vein have largely used expensive per-instance annotation or singular class-level descriptions, but per-instance descriptions are hard to scale and single class descriptions may not be rich enough. Furthermore, these works have used natural-language descriptions exclusively, simple bi-encoders models, and modality or task-specific methods. These approaches have several limitations: text supervision may not always be available or optimal and bi-encoders may only learn coarse relations between inputs and class descriptions. In this work, we present Semsup, a novel approach that uses (1) a scalable multiple description sampling method which improves performance over single descriptions, (2) alternative description formats such as JSON that are easy to generate and outperform text on certain settings, and (3) hybrid lexical-semantic similarity to leverage fine-grained information in class descriptions. We demonstrate the effectiveness of Semsup across four datasets, two modalities, and three generalization settings. For example, across text and image datasets, Semsup increases unseen class generalization accuracy by 15 points on average compared to the closest baseline. Reading List: [ https://docs.google.com/document/d/1S4U2c1IHB5yIHDjQ8yIFLxcKmIbdAm94R1Re3zn… | https://docs.google.com/document/d/1S4U2c1IHB5yIHDjQ8yIFLxcKmIbdAm94R1Re3zn… ] 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