Quick links

CS Department Colloquium Series

Towards Responsible Machine Learning in Societal Systems

Date and Time
Wednesday, April 5, 2023 - 12:30pm to 1:30pm
Location
Computer Science Small Auditorium (Room 105)
Type
CS Department Colloquium Series
Host
Olga Russakovsky and Peter Ramadge

Lydia Liu
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.

To request accommodations for a disability please contact Emily Lawrence, emilyl@cs.princeton.edu, at least one week prior to the event.
 

 

Controlling Large Language Models: Generating (Useful) Text from Models We Don’t Fully Understand

Date and Time
Thursday, March 23, 2023 - 12:30pm to 1:30pm
Location
Computer Science Small Auditorium (Room 105)
Type
CS Department Colloquium Series
Host
Danqi Chen

Ari Holtzman
Generative language models have recently exploded in popularity, with services such as ChatGPT deployed to millions of users. These neural models are fascinating, useful, and incredibly mysterious: rather than designing what we want them to do, we nudge them in the right direction and must discover what they are capable of. But how can we rely on such inscrutable systems?

This talk will describe a number of key characteristics we want from generative models of text, such as coherence and correctness, and show how we can design algorithms to more reliably generate text with these properties. We will also highlight some of the challenges of using such models, including the need to discover and name new and often unexpected emergent behavior. Finally, we will discuss the implications this has for the grand challenge of understanding models at a level where we can safely control their behavior.

Bio: Ari Holtzman is a PhD student at the University of Washington. His research has focused broadly on generative models of text: how we can use them and how can we understand them better. His research interests have spanned everything from dialogue, including winning the first Amazon Alexa Prize in 2017, to fundamental research on text generation, such as proposing Nucleus Sampling, a decoding algorithm used broadly in deployed systems such as the GPT-3 API and academic research. Ari completed an interdisciplinary degree at NYU combining Computer Science and the Philosophy of Language.


To request accommodations for a disability please contact Emily Lawrence, emilyl@cs.princeton.edu, at least one week prior to the event.

Integrating expertise into computational tools for design and media authoring

Date and Time
Wednesday, March 8, 2023 - 12:30pm to 1:30pm
Location
Computer Science Small Auditorium (Room 105)
Type
CS Department Colloquium Series
Host
Andrés Monroy-Hernández

Mackenzie Leake
Finding a good computational representation for a problem allows us to map high level objectives to low level details and select the appropriate set of algorithmic tools. Selecting this representation requires not only computational knowledge but also a deep understanding of the application domain. In this talk I will discuss my work on building design and media authoring tools by combining domain expertise with a wide range of algorithmic techniques. I will describe how this approach helps us to offload tedious steps to computation and guide users’ attention toward the more creative, open-ended decisions. As two different examples of this approach, I will discuss my work on video editing and quilt design tools. I will also discuss future opportunities to combine domain expertise and algorithmic insights to build novel computational tools.

Bio: Mackenzie Leake is a METEOR postdoctoral fellow at MIT CSAIL. She received her PhD and MS in computer science from Stanford University and a BA in computational science and studio art from Scripps College. Her research in human-computer interaction and computer graphics focuses on designing computational tools for various creative domains, including textiles and video. Her research has been supported by Adobe Research, Brown Institute for Media Innovation, and Stanford Enhancing Diversity in Graduate Education (EDGE) fellowships. In 2022 she was named a Rising Star in EECS and a WiGraph Rising Star in Computer Graphics.


To request accommodations for a disability please contact Emily Lawrence, emilyl@cs.princeton.edu, at least one week prior to the event.

Designing Provably Performant Networked Systems

Date and Time
Monday, April 3, 2023 - 12:30pm to 1:30pm
Location
Computer Science Small Auditorium (Room 105)
Type
CS Department Colloquium Series
Host
Ravi Netravali

Venkat Arun
As networked systems become critical infrastructure, their design must reflect their new societal role. Today, we build systems with hundreds of heuristics but often do not understand their inherent and emergent behaviors. I will present a set of tools and techniques to prove performance properties of heuristics running in real-world conditions. Rigorous proofs can not only inspire confidence in our designs, but also give counter-intuitive insights about their performance.

A key theme in our approach is to model uncertainty in systems using non-random, non-deterministic objects that cover a wide range of possible behaviors under a single abstraction. Such models allow us to analyze complex system behaviors using automated reasoning techniques. I will present automated tools to analyze congestion control and process scheduling algorithms. These tools prove performance properties and find counter-examples where widely deployed heuristics fail. I will also show that current end-to-end congestion control algorithms that bound delay cannot avoid starvation and present a method to beamform wireless signals using thousands of antennas.

Bio: Venkat Arun is a PhD candidate at MIT working with Hari Balakrishnan and Mohammad Alizadeh. His work spans internet congestion control, video streaming, privacy-preserving computation, wireless networks, and mobile systems. Across these areas, a unifying theme of his work is to bridge between heuristics that systems use in practice and proofs of how well they work. He believes that rigorous proof combined with automated reasoning will enable us to make networked systems more robust and performant. He has won two ACM SIGCOMM best paper awards and the president of India gold medal.


To request accommodations for a disability please contact Emily Lawrence, emilyl@cs.princeton.edu, at least one week prior to the event.

Self-Supervised Reinforcement Learning

Date and Time
Tuesday, March 21, 2023 - 12:30pm to 1:30pm
Location
Computer Science Small Auditorium (Room 105)
Type
CS Department Colloquium Series
Host
Karthik Narasimhan

Benjamin Eysenbach
Reinforcement learning (RL) promises to harness the power of machine learning to solve sequential decision making problems, with the potential to enable applications ranging from robotics to chemistry. However, what makes the RL paradigm broadly applicable is also what makes it challenging: only limited feedback is provided for learning to select good actions. In this talk, I will discuss how we have made headway of this challenge by designing a class of self-supervised RL methods, ones that can learn skills for acting using unsupervised (reward-free) experience. These skill learning methods are practically-appealing and have since sparked a vibrant area of research. I will also share how we have answered some open theoretical questions in this area.

Bio: Benjamin Eysenbach a final-year PhD student at Carnegie Mellon University. His research has developed machine learning algorithms for sequential decision making. His algorithms not only achieve a high degree of performance, but also carry theoretical guarantees, are typically simpler than prior methods, and draw connections between many areas of ML and CS. Ben is the recipient of the NSF and Hertz graduate fellowships. Prior to the PhD, he was a resident at Google Research and studied math as an undergraduate at MIT.


To request accommodations for a disability please contact Emily Lawrence, emilyl@cs.princeton.edu, at least one week prior to the event.

Programming Distributed Systems

Date and Time
Thursday, April 6, 2023 - 12:30pm to 1:30pm
Location
Computer Science Small Auditorium (Room 105)
Type
CS Department Colloquium Series
Host
Andrew Appel

Mae Milano
Our interconnected world is increasingly reliant on distributed systems of unprecedented scale, serving applications which must share state across the globe. And, despite decades of research, we're still not sure how to program them!  In this talk, I'll show how to use ideas from programming languages to make programming at scale easier, without sacrificing performance, correctness, or expressive power in the process.  We'll see how slight tweaks to modern imperative programming languages can provably eliminate common errors due to replica consistency or concurrency---with little to no programmer effort.  We'll see how new language designs can unlock new systems designs, yielding both more comprehensible protocols and better performance.  And we'll conclude by imagining together the role that a new cloud-centric programming language could play in the next generation of distributed programs.

Bio: Mae Milano is a postdoctoral scholar at UC Berkeley working at the intersection of Programming Languages, Distributed Systems, and Databases.  Her work has appeared at top-tier venues including PLDI, OOPSLA, POPL, VLDB, and TOCS, and has attracted the attention of the Swift language team. She is a recipient of the NDSEG Fellowship, has won several awards for her writing and service, and is a founding member of the Computing Connections Fellowship's selection committee (https://computingconnections.org/).


To request accommodations for a disability please contact Emily Lawrence, emilyl@cs.princeton.edu, at least one week prior to the event.

Hardware-aware Algorithms for Efficient Machine Learning

Date and Time
Thursday, March 2, 2023 - 12:30pm to 1:30pm
Location
Computer Science Small Auditorium (Room 105)
Type
CS Department Colloquium Series
Host
Jia Deng

Tri Dao
Machine learning (ML) models training will continue to grow to consume more cycles, their inference will proliferate on more kinds of devices, and their capabilities will be used on more domains. Some goals central to this future are to make ML models efficient so they remain practical to train and deploy, and to unlock new application domains with new capabilities. We describe some recent developments in hardware-aware algorithms to improve the efficiency-quality tradeoff of ML models and equip them with long context. In the first half, we focus on structured sparsity, a natural approach to mitigate the extensive compute and memory cost of large ML models. We describe a line of work on learnable fast transforms which, thanks to their expressiveness and efficiency, yields some of the first sparse training methods to speed up large models in wall-clock time (2x) without compromising their quality. In the second half, we focus on efficient Transformer training and inference for long sequences. We describe FlashAttention, a fast and memory-efficient algorithm to compute attention with no approximation. By careful accounting of reads/writes between different levels of memory hierarchy, FlashAttention is 2-4x faster and uses 10-20x less memory compared to the best existing attention implementations, allowing us to train higher-quality Transformers with 8x longer context. FlashAttention is now widely used in some of the largest research labs and companies, in just 6 months after its release. We conclude with some exciting directions in ML and systems, such as software-hardware co-design, structured sparsity for scientific AI, and long context for new AI workflows and modalities.

Bio: Tri Dao is a PhD student in Computer Science at Stanford, co-advised by Christopher Ré and Stefano Ermon. He works at the interface of machine learning and systems, and his research interests include sequence models with long-range memory and structured matrices for compact deep learning models. His work has received the ICML 2022 Outstanding paper runner-up award.


To request accommodations for a disability please contact Emily Lawrence, emilyl@cs.princeton.edu, at least one week prior to the event.

The Design of a General-Purpose Distributed Execution System

Date and Time
Wednesday, March 1, 2023 - 12:30pm to 1:30pm
Location
Computer Science Small Auditorium (Room 105)
Type
CS Department Colloquium Series
Host
Wyatt Lloyd

Stephanie Wang
Scaling applications with distributed execution has become the norm. With the rise of big data and machine learning, more and more developers must build applications that involve complex and data-intensive distributed processing.

In this talk, I will discuss the design of a general-purpose distributed execution system that can serve as a common platform for such applications. Such a system offers two key benefits: (1) common system functionality such as distributed resource management can be shared across different application domains, and (2) by building on the same platform, applications across domains can easily interoperate.

First, I will introduce the distributed futures interface, a powerful yet expressive distributed programming abstraction for remote execution and memory. Second, I will introduce ownership, an architecture for distributed futures systems that simultaneously provides horizontal scalability, low latency, and fault tolerance. Finally, I will present Exoshuffle, a large-scale shuffle system that builds on distributed futures and ownership to match the speed and reliability of specialized data processing frameworks while using an order of magnitude less code. These works have reached a broad audience through Ray, an open-source distributed futures system for Python that has more than 23,000 GitHub stars and that has been used to train ChatGPT and to break the world record for CloudSort.

Bio: Stephanie Wang is a final-year PhD student at UC Berkeley, advised by Professor Ion Stoica. She is interested in distributed systems, with current focus on problems in cloud computing and fault tolerance. She is a co-creator and committer of the popular open-source project Ray for distributed Python. Stephanie has received the UC Berkeley Chancellor’s Fellowship, a Distinguished Artifact Award at SOSP’19, and was selected for Rising Stars in EECS in 2021.


To request accommodations for a disability please contact Emily Lawrence, emilyl@cs.princeton.edu, at least one week prior to the event.

Designing Formally Correct Intermittent Systems

Date and Time
Monday, February 27, 2023 - 12:30pm to 1:30pm
Location
Computer Science Small Auditorium (Room 105)
Type
CS Department Colloquium Series
Host
Amit Levy

Milijana Surbatovich
"Extreme edge computing" is an emerging computing paradigm targeting application  domains like medical wearables, disaster-monitoring tiny satellites, or smart infrastructure. This paradigm brings sophisticated sensing and data processing  into an embedded device's deployment environment, enabling computing in environments  that are too harsh, inaccessible, or dense to support frequent communication with a central server. Batteryless, energy harvesting devices (EHDs) are key to enabling extreme edge computing; instead of using batteries, which may be too costly or even impossible to replace, they can operate solely off energy collected from their environment. However, harvested energy is typically too weak to power a device continuously, causing frequent, arbitrary power failures that break  traditional software and make correct programming difficult. Given the high assurance requirements of the envisioned application domains, EHDs must execute software without bugs that could render the device inoperable or leak sensitive information. While researchers have developed intermittent systems to support programming EHDs, they rely on informal, undefined correctness notions that preclude proving such necessary correctness and security properties.

My research lays the foundation for designing formally correct intermittent systems that provide correctness guarantees. In this talk, I show how existing correctness notions are insufficient, leading to unaddressed bugs. I then present the first formal model of intermittent execution, along with correctness definitions for important memory consistency and timing properties. I use these definitions to design and implement both the language abstractions that programmers can use to specify their desired properties and the enforcement mechanisms that uphold them. Finally, I discuss my future research directions in intermittent system security and leveraging formal methods for full-stack correctness reasoning. 

Bio: Milijana Surbatovich is a PhD Candidate in the Electrical and Computer Engineering Department at Carnegie Mellon University, co-advised by Professors Brandon Lucia and Limin Jia. Her research interests are in applied formal methods, programming languages, and systems for intermittent computing and non-traditional computing platforms broadly. She is excited by research problems that require reasoning about correctness and security across the architecture, system, and language stack. She was awarded CMU's CyLab Presidential Fellowship in 2021 and was selected as a 2022 Rising Star in EECS. Previously, she received an MS in ECE from CMU in 2020 and a BS in Computer Science from the University of Rochester in 2017. 


To request accommodations for a disability please contact Emily Lawrence, emilyl@cs.princeton.edu, at least one week prior to the event.

Detecting COVID-19 with Genomic Sequencing: From bench to vending machine

Date and Time
Tuesday, February 21, 2023 - 12:30pm to 1:30pm
Location
Computer Science Small Auditorium (Room 105)
Type
CS Department Colloquium Series
Speaker
Dr. Eleazar Eskin , from UCLA
Host
Ben Raphael

At UCLA we developed one of the only novel technologies for COVID-19 diagnostic testing that was deployed on a large scale.  The assay, which we named SwabSeq, performs genomic sequencing of pooled samples tagged with sample-specific molecular barcodes and then uses computational approaches to deconvolve the pooled samples into individual diagnoses, enabling the testing of thousands of nasal or saliva samples for SARS-CoV-2 RNA in a single run without the need for RNA extraction.   The efficiency of SwabSeq has enabled a small facility with a handful of staff to perform over 1,500,000 tests, with an analytical sensitivity and specificity comparable to or better than traditional qPCR test with turnaround times of less than 24 h. SwabSeq could be rapidly adapted for the detection of other pathogens.

 

Bio:  Dr. Eleazar Eskin is the founding Chair of the Department of Computational Medicine at UCLA.  He is also a Professor in the Computational Medicine, Computer Science and Human Genetics departments.  His research focuses on developing computational methods for the analysis of genetic variation and discovering the genetic basis of human disease.  He was a founding faculty of multiple academic programs at UCLA including the Bioinformatics Ph.D. Program, the Undergraduate Minor in Bioinformatics, the Bruins in Genomics Summer Research Program and the Computational Genomics Summer Institute.  He is a Fellow of the International Society of Computational Biology and a Alfred P Sloan Foundation Research Fellow. To learn more about Eleazar's work please visit: http://web.cs.ucla.edu/~eeskin/

 

To request accommodations for a disability, please contact Michael Estepp at mestepp@cs.princeton.edu at least one week prior to the event.

Follow us: Facebook Twitter Linkedin