- talks - lists.cs.princeton.edu

CS Colloquium Series: week of March 14-18
by Emily C. Lawrence 11 Mar '22

11 Mar '22

Here is the full list of CS Colloquium talks for next week. All talks will be recorded. ~~~~~ Speaker: Kostis Kaffes, Stanford University Date: Monday, March 14, 2022 Time: 12:30pm EST Location: Zoom Webinar Host: Ravi Netravali Event page: [ https://www.cs.princeton.edu/events/26171 | https://www.cs.princeton.edu/events/26171 ] Please register here: [ https://princeton.zoom.us/webinar/register/WN_Nn4iOb-mRqSEJbGsbzg5uQ | https://princeton.zoom.us/webinar/register/WN_Nn4iOb-mRqSEJbGsbzg5uQ ] Title: Solving the Cloud Efficiency Crisis with Fast and Accessible Scheduling Abstract: Operating systems (OS) specialization is necessary as the one-size-fits-all approach of fundamental OS operations such as scheduling is incompatible with today's diverse application landscape. Such specialization can improve application performance and cloud platform efficiency by an order of magnitude or more. Towards this goal, I will first discuss Shinjuku, a specialized OS that supports an order of magnitude higher load and lower tail latency than state-of-the-art systems by enabling better scheduling. Shinjuku leverages hardware support for virtualization to preempt as often as every 5 microseconds and disproves the conventional wisdom that interrupts are incompatible with microsecond timescales. Then, I will present Syrup, a framework that enables everyday application developers to specify custom scheduling policies easily and safely deploy them across different layers of the stack over existing operating systems like Linux, bringing the benefits of specialized scheduling to everyone. For example, Syrup allowed us to implement policies that previously required specialized dataplanes in less than 20 lines of code and improve the performance of an in-memory database by 8x without needing any application modification. Bio: Kostis Kaffes is a final-year Ph.D. candidate in Electrical Engineering at Stanford University, advised by Christos Kozyrakis. He is broadly interested in computer systems, cloud computing, and scheduling. His thesis focuses on end-host, rack-scale, and cluster-scale scheduling for microsecond-scale tail latency with the goal of improving efficiency in the cloud. Recently, he has been looking for ways to make it easier to implement and deploy custom scheduling policies across different layers of the stack. Kostis's research has been supported by a Facebook Research Award and various scholarships and fellowships from Stanford, A.G. Leventis Foundation, and Gerondelis Foundation. Prior to Stanford, he received his undergraduate degree in Electrical and Computer Engineering from the National Technical University of Athens in Greece. ~~~~~ Speaker: Gilbert Bernstein, University of California, Berkeley and MIT Date: Tuesday, March 15, 2022 Time: 12:30pm EST Location: CS 105 Host: Felix Heide Event page: [ https://www.cs.princeton.edu/events/26172 | https://www.cs.princeton.edu/events/26172 ] This talk will be live-streamed at [ https://mediacentrallive.princeton.edu/ | https://mediacentrallive.princeton.edu/ ] Title: High-Performance Languages for Visual Computing Abstract: Computer Graphics and Visual Computing problems challenge us with a near inexhaustible demand for more resolution and scale in order to simulate the climate, reconstruct 3d environments, train neural networks, and produce games & movies. Building such applications requires integrating disciplinary expertise (e.g. physics, numerical methods, geometry and custom hardware) into a single system. However, abstraction barriers are regularly discarded in the name of higher-performance, leading to code that must be written and maintained by super-experts—programmers who simultaneously possess deep knowledge of all relevant disciplines. Programming languages, especially Domain Specific Languages (DSLs) are perhaps the most promising approach to recovering a separation of concerns in such high-performance systems. In this talk, I will first describe my work on DSLs to enable parallel portability of physical simulation and optimization programs, including the use of relational algebra and automatic differentiation to structure these problem domains. Then I will discuss more recent work on “horizontal DSLs” designed to dig into specific sub-problems: maximizing utilization of novel hardware accelerators, formally verifying optimizations of tensor programs, and extending automatic differentiation to correctly handle discontinuities in inverse rendering and simulation problems. Bio: Gilbert Bernstein is a Postdoctoral Scholar at the University of California, Berkeley and MIT CSAIL, working with Professor Jonathan Ragan-Kelley. His research lies in Computer Graphics and Programming Languages, especially the design of high-performance domain specific languages for numeric computing applications such as physical simulation, optimization and inverse problems. His work spans the gamut from user interfaces, to differentiable programming, parallel-portability, and new hardware design languages. His work has been published at SIGGRAPH, POPL, PLDI, & OOPSLA, as well as being incorporated into products at Adobe, Autodesk, and Disney. He holds a Ph.D. in Computer Science from Stanford University, where he was advised by Pat Hanrahan. ~~~~~ Speaker: Adam Yala, Massachusetts Institute of Technology Date: Wednesday, March 16, 2022 Time: 12:30pm EST Location: CS 105 Host: Karthik Narasimhan Event page: [ https://www.cs.princeton.edu/events/26173 | https://www.cs.princeton.edu/events/26173 ] This talk will be live-streamed at [ https://mediacentrallive.princeton.edu/ | https://mediacentrallive.princeton.edu/ ] Title: Seeing into the future: Machine learning methods for personalized screening Abstract: While AI has the potential to transform patient care: improving outcomes, reducing costs and eliminating health disparities. However, the development of equitable clinical AI models and their translation to hospitals remains difficult. From a computational perspective, these tools must deliver consistent performance across diverse populations while learning from biased and scarce data. In this talk, I will discuss approaches addressing the above challenges in three areas: 1) cancer risk assessment from imaging, 2) personalized screening policy design and 3) private data sharing through neural obfuscation. I’ve demonstrated that these clinical models offer significant improvements over the current standard of care across globally diverse patient populations. Bio: Adam Yala is a PhD student in Electrical Engineering and Computer Science at MIT. He is a member of MIT Jameel Clinic for AI and Health and a member of CSAIL. His research focuses on developing machine learning methods for personalized medicine and translating them to clinical care. Adam's tools have been deployed at multiple health systems around the world and his research has been featured in the Washington Post, New York Times, Boston Globe and Wired. ~~~~~ Speaker: Pang Wei Koh, Stanford University Date: Thursday, March 17, 2022 Time: 12:30pm EST Location: CS 105 Host: Sanjeev Arora Event page: [ https://www.cs.princeton.edu/events/26174 | https://www.cs.princeton.edu/events/26174 ] This talk will be live-streamed at [ https://mediacentrallive.princeton.edu/ | https://mediacentrallive.princeton.edu/ ] Title: Reliable machine learning in the wild Abstract: Machine learning systems are widely deployed today, but they are unreliable. They can fail – and with catastrophic consequences – on subpopulations of the data, such as particular demographic groups, or when deployed in different environments from what they were trained on. In this talk, I will describe our work towards building reliable machine learning systems that are robust to these failures. First, I will show how we can use influence functions to understand the predictions and failures of existing models through the lens of their training data. Second, I will discuss the use of distributionally robust optimization to train models that perform well across all subpopulations. Third, I will describe WILDS – a benchmark of in-the-wild distribution shifts spanning applications such as pathology, conservation, remote sensing, and drug discovery – and show how current state-of-the-art methods, which perform well on synthetic distribution shifts, still fail to be robust on these real-world shifts. Finally, I will describe our work on building more reliable COVID-19 models, using anonymized cellphone mobility data, to inform public health policy; this is a challenging application as the underlying environment is often changing and there is substantial heterogeneity across demographic subpopulations. Bio: Pang Wei Koh is a PhD student at Stanford, advised by Percy Liang. He studies the theory and practice of building reliable machine learning systems. His research has been published in Nature and Cell, featured in media outlets such as The New York Times and The Washington Post, and recognized by best paper awards at ICML and KDD, a Meta Research PhD fellowship, and the Kennedy Prize for best honors thesis at Stanford. Prior to his PhD, he was the 3rd employee and Director of Partnerships at Coursera.

1 1

Ryan Amos will present his FPO "Consumer Protection on the Web with Longitudinal Web Crawls and Analysis" on March 17, 2022 at 3pm in CS 402.
by Nicki Mahler 10 Mar '22

10 Mar '22

Ryan Amos will present his FPO "Consumer Protection on the Web with Longitudinal Web Crawls and Analysis" on March 17, 2022 at 3pm in CS 402. The members of his committee are as follows: Examiners: Edward Felten (adviser), Prateek Mittal (adviser), and Arvind Narayanan; Readers: Jonathan Mayer and Joe Calandrino (Federal Trade Commission) A copy of his thesis is available upon request. Please email gradinfo(a)cs.princeton.edu if you would like a copy of the thesis. Everyone is invited to attend his talk. Abstract: The world wide web has brought with it new consumer protection hazards, such as deceptive reviews and online tracking. While many academics have studied consumer protection on the web at specific points in time, we approach this problem from a longitudinal perspective, exploring how consumers' rights to privacy and to be informed have been impacted by the web. Our work highlights the key role in study of consumer protection issues played by longitudinal analyses and longitudinal data collection -- data collected over repeated, time-spaced passes. We investigate consumer protection issues on the web through longitudinal studies in two landscapes: website privacy policies and reviews on Yelp. We approach both problems by collecting data with automated, repeated visits to the websites of interest to collect large scale datasets. In our study of privacy policies, we aggregate Internet Archive's crawls to perform longitudinal collection, and in our online reviews study, we crawl the data ourselves. We collected 1M privacy policies spanning 22 years and 12.5M reviews over 11 months. We used our data to study the evolution of privacy policies raising concerns with rights to privacy and information. We find gaps in disclosure of privacy-related practices. We show declining readability over the long term, doubling in length and becoming more complex. We show disparities in website-reported and independently-observed tracking. In our study of online reviews we raise concerns with the right to be informed. We present the first study of "reclassification," wherein a platform changes its filtering decision for a review. We find that reviews routinely move between Yelp's two main classifier classes ("Recommended" and "Not Recommended"), up to five reclassifications on a single review. We identify demographic disparities in review prevalence and filtering decisions. By showing phenomena that cannot be studied without longitudinal data collection and analysis, we emphasize the importance of longitudinal study for consumer protection issues online. Our work helps lay the groundwork for future work on these issues through our software and data releases, easing the pathway for future researchers.

1 1

ECE SEMINAR Heng Yang, MIT 3/16/2022
by Emily C. Lawrence 10 Mar '22

10 Mar '22

ECE SEMINAR Speaker: Heng Yang, MIT Title: Certifiable Outlier-Robust Geometric Perception: Robots that See through the Clutter with Confidence Day: March 16, 2022 Time: 4:30pm Location: B205 Engineering Quadrangle Host: Jaime Fernández Fisac Abstract: Geometric perception is the task of estimating geometric models (e.g., object pose and 3D structure) from sensor measurements and priors (e.g., point clouds and neural network detections). Geometric perception is a fundamental building block for robotics applications ranging from intelligent transportation to space autonomy. The ubiquitous existence of outliers -measurements that tell no or little information about the models to be estimated- makes it theoretically intractable to perform estimation with guaranteed optimality. Despite this theoretical intractability, safety-critical robotics applications still demand trustworthiness and performance guarantees on perception algorithms. In this talk, I present certifiable outlier-robust geometric perception, a new paradigm to design tractable algorithms that enjoy rigorous performance guarantees, i.e., they return an optimal estimate with a certificate of optimality for a majority of problem instances, but declare failure and provide a measure of suboptimality for worst-case instances. Particularly, I present two general-purpose algorithms in the certifiable perception toolbox: (i) an estimator that uses graph theory to prune gross outliers and leverages graduated non-convexity to compute the optimal model estimate with high probability of success, and (ii) a certifier that employs sparse semidefinite programming (SDP) relaxation and a novel SDP solver to endow the estimator with an optimality certificate or escape local minima otherwise. The estimator is fast and robust against up to 99% random outliers in practical perception applications, and the certifier can compute high-accuracy optimality certificates for large-scale problems beyond the reach of existing SDP solvers. I showcase certifiable outlier-robust perception on robotics applications such as scan matching, satellite pose estimation, and vehicle pose and shape estimation. I conclude by remarking three opportunities arising from certifiable perception: to speedup online global optimization by offline learning from data; to enable safe learning-based perception by bridging certifiable estimation with deep representation learning; and to couple and unify perception with action towards trustworthy autonomy. Bio: Heng Yang is a final-year Ph.D. candidate in the Laboratory for Information & Decision Systems and the Department of Mechanical Engineering at the Massachusetts Institute of Technology (MIT), working with Prof. Luca Carlone. He holds a B.S. degree from Tsinghua University and an S.M. degree from MIT, both in Mechanical Engineering. His research interests include large-scale convex optimization, semidefinite relaxation, robust estimation, and machine learning, applied to robotics and trustworthy autonomy. His work includes developing certifiable outlier-robust machine perception algorithms, large-scale semidefinite programming solvers, and self-supervised geometric perception frameworks. Heng Yang is a recipient of the Best Paper Award in Robot Vision at the 2020 IEEE International Conference on Robotics and Automation (ICRA), a Best Paper Award Honorable Mention from the 2020 IEEE Robotics and Automation Letters (RA-L), and a Best Paper Award Finalist at the 2021 Robotics: Science and Systems (RSS) conference. He is a Class of 2021 RSS Pioneer.

1 0

UPDATE: Ankit Goyal will present his Pre FPO on Wednesday, March 16 at 11am on Zoom & CS 302
by Louis Riehl 09 Mar '22

09 Mar '22

UPDATE: Ankit Goyal will present his Pre FPO on Wednesday, March 16 at 11am on Zoom and in CS 302 Zoom link: https://princeton.zoom.us/j/95476811864 Committee members: Jia Deng (examiner, adviser), Szymon Rusinkiewicz (examiner), Danqi Chen (examiner), Karthik Narasimhan (reader), Vladlen Koltun (reader, Apple) Abstract: Embodied Intelligent Agents are systems that have physical embodiment and can change the physical world to accomplish various goals. To act successfully, agents must have three capabilities. First, they must be able to perceive the 3D environment. Second, they must be able to reason about the 3D space. And third, they must be able to talk to humans about the 3D world. In this talk, I will present our efforts in building systems and benchmarks to improve embodied agents across all three capabilities

1 0

Ankit Goyal will present his Pre FPO on Wednesday, March 16 at 11am on Zoom.
by Louis W. Riehl 09 Mar '22

09 Mar '22

Ankit Goyal will present his Pre FPO on Wednesday, March 16 at 11am on Zoom. Zoom link: [ https://princeton.zoom.us/j/95476811864 | https://princeton.zoom.us/j/95476811864 ] Committee members: Jia Deng (examiner, adviser), Szymon Rusinkiewicz (examiner), Danqi Chen (examiner), Karthik Narasimhan (reader), Vladlen Koltun (reader, Apple) Abstract: Embodied Intelligent Agents are systems that have physical embodiment and can change the physical world to accomplish various goals. To act successfully, agents must have three capabilities. First, they must be able to perceive the 3D environment. Second, they must be able to reason about the 3D space. And third, they must be able to talk to humans about the 3D world. In this talk, I will present our efforts in building systems and benchmarks to improve embodied agents across all three capabilities

1 0

Alejandro Newell will present his FPO "Learning to Solve Structured Vision Problems" on Tuesday, March 15, 2022 at 10 AM at Friend 125 and on Zoom (https://princeton.zoom.us/j/7228079201)
by Grad Info 08 Mar '22

08 Mar '22

Alejandro Newell will present his FPO "Learning to Solve Structured Vision Problems" on Tuesday, March 15, 2022 at 10AM at Friend 125 and on Zoom (https://princeton.zoom.us/j/7228079201). Location: Friend 125 & https://princeton.zoom.us/j/7228079201 The members of Alejandro's committee are as follows: Examiners: Jia Deng (advisor), Adam Finkelstein, Felix Heide Readers: Olga Russakovsky, Szymon Rusinkiewicz A copy of his thesis is available upon request. Please email gradinfo(a)cs.princeton.edu <mailto:gradinfo@cs.princeton.edu> if you would like a copy of the thesis. Everyone is invited to attend his talk. Abstract follows below: We want computer vision models to understand the rich world captured in images and video. This requires not just recognizing objects, but identifying their relationships and interactions. Combining contributions in both neural architecture and loss design, we expand the capacity of convolutional networks to express such interactions and solve a broad range of structured computer vision tasks. We first introduce a convolutional network architecture for dense per-pixel prediction. We show how intermediate supervision and repeated processing across feature scales lead to better network performance, referring to the architecture as a “stacked hourglass” network based on the successive steps of pooling and up sampling during inference. We benchmark on the task of human pose estimation achieving state-of-the-art performance. Next, we introduce associative embedding, a method for supervising networks to solve detection and grouping tasks. A number of problems can be framed in this manner including multi-person pose estimation, instance segmentation, and multi-object tracking. Usually these problems are addressed with multi-stage pipelines, instead we train a model to simultaneously output detections and group assignments. We can then extend the use of associative embeddings to define arbitrary graphs. We demonstrate how to supervise embeddings such that a model both detects the objects in a scene and defines semantic relationships between pairs of objects. Finally, we perform an investigation of self-supervision methods. Recent self-supervised losses rely on a similar learning signal to the loss we leverage in our as-sociative embedding work. But it is unclear how useful these losses are for general purpose visual feature pretraining. We investigate what factors play a role in the utility of such pretraining by evaluating self-supervised algorithms across a comprehensive array of synthetic datasets and downstream tasks. Our experiments highlight how self-supervision can be more or less useful depending on the amount of labeled data, the complexity of the data, and the target downstream task. Together the work in this thesis shows how to build and train better models while providing insights into what steps lead to the best performance across a wide variety of computer vision tasks.

1 0

ECE SEMINAR Christopher Torng, Stanford University 3/7/2022
by Emily C. Lawrence 03 Mar '22

03 Mar '22

ECE SEMINAR Speaker: Christopher Torng, Stanford University Title: Accelerating Chip-Building Design Cycles for Future Generations of Computing Day: March 7, 2022 Time: 4:30 pm Zoom Link: Please sign up at [ https://princeton.zoom.us/meeting/register/tJUtcuuqrTwvHNWG3rFwtpcYlfsC8LtS… | https://princeton.zoom.us/meeting/register/tJUtcuuqrTwvHNWG3rFwtpcYlfsC8LtS… ] Host: David Wentzlaff Abstract: The chip building industry is a major cornerstone of the global economy. As a result, addressing the causes behind a multi-year global chip shortage is important for both near and long term futures. Unfortunately, one major challenge is that it is difficult to produce high-quality designs quickly and at low cost using traditional hardware design flows. This means that the industry wastes valuable fabrication slots learning painful design lessons rather than meeting economic demands. My research focuses on building new architectures, systems, and design tools to accelerate chip building design cycles for future generations of computing systems. To support this goal, my research spans across the computing stack, ranging from applications, compilers, architectures, and down to chip implementation. In this talk, I will first present a set of vertically integrated techniques (compiler, architecture, and VLSI) that significantly reduces the design effort for extremely fine-grain power control in spatial architectures. Next, I will introduce my work on a new generation of open and agile hardware flow tools that leverage modern programming language features to increase code reuse in physical design. Finally, I will discuss recent work on Amber SoC, a coarse-grained reconfigurable array designed with an end-to-end agile accelerator-compiler co-designed flow. I will conclude with my future directions in supporting chip building for the next generation of computing. Bio: Christopher Torng is a postdoctoral researcher at Stanford University. He received his Ph.D. degree, M.S. degree, and B.S degree (2019, 2016, 2012) in Electrical and Computer Engineering from Cornell University. His projects target the development of architectures and tools to accelerate building chips and complex hardware systems. His tools have achieved use across multiple universities to support over ten academic tapeouts in technologies ranging from 180nm to 16nm. His activities have resulted in a selection as a Rising Star in Computer Architecture (2018) by Georgia Tech and an IEEE MICRO Top Pick from Hot Chips (2018).

1 0

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