- talks - lists.cs.princeton.edu

Reminder: Ellen Vitercik tomorrow at 12:30pm EST
by Emily C. Lawrence 21 Feb '21

21 Feb '21

CS Colloquium Speaker Ellen Vitercik, Carnegie Mellon University Host: Sanjeev Arora Date: Monday, February 22 Time: 12:30pm EST Event page: [ https://www.cs.princeton.edu/events/26042 | https://www.cs.princeton.edu/events/26042 ] Register for the Zoom webinar here: [ https://princeton.zoom.us/webinar/register/WN_gbMN0OUJQRainEiRTS4VYQ | https://princeton.zoom.us/webinar/register/WN_gbMN0OUJQRainEiRTS4VYQ ] Title: Integrating Machine Learning into Algorithm Design Abstract: An important property of those algorithms that are typically used in practice is broad applicability—the ability to solve problems across diverse domains. However, the default, out-of-the-box performance of these algorithms can be unsatisfactory, with slow runtime, poor solution quality, and even negative long-term social ramifications. In practice, there is often ample data available about the types of problems that an algorithm will be run on, data that can potentially be harnessed to fine-tune the algorithm’s performance. We therefore need principled approaches for using this data to obtain strong application-specific performance guarantees. In this talk, I will give an overview of my research that provides practical methods built on firm theoretical foundations for incorporating machine learning and optimization into the process of algorithm design, selection, and configuration. I will describe my contributions across several diverse domains, including integer programming, clustering, mechanism design, and computational biology. As I will demonstrate, these seemingly disparate areas are connected by overarching structure which implies broadly-applicable guarantees. Bio: Ellen Vitercik is a PhD student at Carnegie Mellon University where she is co-advised by Maria-Florina Balcan and Tuomas Sandholm. Her research revolves around artificial intelligence, algorithm design, and the interface between economics and computation, with a particular focus on machine learning theory. Among other honors, she is a recipient of the Exemplary Artificial Intelligence Track Paper Award at EC'19, the Best Presentation by a Student or Postdoctoral Researcher Award at EC'19, the NSF Graduate Research Fellowship, the IBM PhD Fellowship, the Fellowship in Digital Health from CMU's Center for Machine Learning and Health, and the Teaching Assistant of the Year Award from CMU's Machine Learning Department. This talk will be recorded.

1 0

CS Distinguished Colloquium Speaker Christos Papadimitriou, today at 12:30
by Emily C. Lawrence 19 Feb '21

19 Feb '21

CS Distinguished Colloquium Speaker Christos Papadimitriou, Columbia University Host: Sanjeev Arora Date: Friday, February 19 Time: 12:30pm EST Event page: https://www.cs.princeton.edu/events/26038 Register for Zoom webinar here: https://princeton.zoom.us/webinar/register/WN_oQAMcljNQbWYTJcaDmTKDQ Title: Language, Brain, and Computation Abstract: How does the brain beget the mind? How do molecules, cells and synapses effect reasoning, intelligence, language? Despite dazzling progress in experimental neuroscience, as well as in cognitive science at the other extreme of scale, we do not seem to be making progress in the overarching question -- the gap is huge and a completely new approach seems to be required. As Richard Axel recently put it: "We don't have a logic for the transformation of neural activity into thought [...]." What kind of formal system would qualify as this "logic"? I will introduce the Assembly Calculus, a computational system whose basic data structure is the assembly -- assemblies are large populations of neurons representing concepts, words, ideas, episodes, etc., The Assembly Calculus is biologically plausible in the following two orthogonal senses: Its primitives are properties of assemblies observed in experiments or useful for explaining experiments, and can be provably (through both mathematical proof and simulations) "compiled down" to the activity of neurons and synapses. I will also explain why the Assembly Calculus is especially powerful in exploring how language happens in the brain. Bio: Christos Harilaos Papadimitriou is the Donovan Family Professor of Computer Science at Columbia University. Before joining Columbia in 2017, he was a professor at UC Berkeley for the previous 22 years, and before that he taught at Harvard, MIT, NTU Athens, Stanford, and UCSD. He has written five textbooks and many articles on algorithms and complexity, and their applications to optimization, databases, control, AI, robotics, economics and game theory, the Internet, evolution, and the brain. He holds a PhD from Princeton (1976), and eight honorary doctorates, including from ETH, University of Athens, EPFL, and Univ. de Paris Dauphine. He is a member of the National Academy of Sciences of the US, the American Academy of Arts and Sciences, and the National Academy of Engineering, and he has received the Knuth prize, the Go"del prize, the von Neumann medal, as well as the 2018 Harvey Prize by Technion. In 2015 the president of the Hellenic republic named him commander of the order of the Phoenix. He has also written three novels: “Turing ,” “Logicomix” and his latest “Independence.”

1 0

Reminder: Christos Papadimitriou, tomorrow at 12:30
by Emily C. Lawrence 18 Feb '21

18 Feb '21

CS Distinguished Colloquium Speaker Christos Papadimitriou, Columbia University Host: Sanjeev Arora Date: Friday, February 19 Time: 12:30pm EST Event page: https://www.cs.princeton.edu/events/26038 Register for Zoom webinar here: https://princeton.zoom.us/webinar/register/WN_oQAMcljNQbWYTJcaDmTKDQ Title: Language, Brain, and Computation Abstract: How does the brain beget the mind? How do molecules, cells and synapses effect reasoning, intelligence, language? Despite dazzling progress in experimental neuroscience, as well as in cognitive science at the other extreme of scale, we do not seem to be making progress in the overarching question -- the gap is huge and a completely new approach seems to be required. As Richard Axel recently put it: "We don't have a logic for the transformation of neural activity into thought [...]." What kind of formal system would qualify as this "logic"? I will introduce the Assembly Calculus, a computational system whose basic data structure is the assembly -- assemblies are large populations of neurons representing concepts, words, ideas, episodes, etc., The Assembly Calculus is biologically plausible in the following two orthogonal senses: Its primitives are properties of assemblies observed in experiments or useful for explaining experiments, and can be provably (through both mathematical proof and simulations) "compiled down" to the activity of neurons and synapses. I will also explain why the Assembly Calculus is especially powerful in exploring how language happens in the brain. Bio: Christos Harilaos Papadimitriou is the Donovan Family Professor of Computer Science at Columbia University. Before joining Columbia in 2017, he was a professor at UC Berkeley for the previous 22 years, and before that he taught at Harvard, MIT, NTU Athens, Stanford, and UCSD. He has written five textbooks and many articles on algorithms and complexity, and their applications to optimization, databases, control, AI, robotics, economics and game theory, the Internet, evolution, and the brain. He holds a PhD from Princeton (1976), and eight honorary doctorates, including from ETH, University of Athens, EPFL, and Univ. de Paris Dauphine. He is a member of the National Academy of Sciences of the US, the American Academy of Arts and Sciences, and the National Academy of Engineering, and he has received the Knuth prize, the Go"del prize, the von Neumann medal, as well as the 2018 Harvey Prize by Technion. In 2015 the president of the Hellenic republic named him commander of the order of the Phoenix. He has also written three novels: “Turing ,” “Logicomix” and his latest “Independence.”

1 0

Gregory Gundersen will present his Pre FPO "Tractable Inference for Latent Variable Models with Applications to Science and Engineering" on Tuesday, February 23, 2021 at 3:30pm via Zoom.
by Nicki Mahler 16 Feb '21

16 Feb '21

Gregory Gundersen will present his Pre FPO " Tractable Inference for Latent Variable Models with Applications to Science and Engineering" on Tuesday, February 23, 2021 at 3:30pm via Zoom. Zoom link: [ https://princeton.zoom.us/j/96395535605 | https://princeton.zoom.us/j/96395535605 ] Committee: Barbara Engelhardt, Ryan Adams, and Jonathan Pillow (examiners); and Tom Griffiths and Brandon Stewart (readers) Title: Tractable Inference for Latent Variable Models with Applications to Science and Engineering Abstract: Latent variables allow practitioners to encode assumptions into their statistical models. A latent variable might, for example, represent an unobserved covariate, measurement error, or a missing class label. Inference is challenging because one must account for the conditional dependence structure induced by these variables, and marginalization is often intractable. In this talk, I present several practical algorithms for inferring latent structure in models used in neuroscience, computational biology, and time-series analysis. First, I present a family of nonlinear dimension-reduction models that use random features to support non-Gaussian data likelihoods. Approximating a nonlinear relationship between the latent variables and observations with a function that is linear with respect to random features induces closed-form gradients of the posterior distribution with respect to the latent variables. This allows for gradient-based nonlinear dimension-reduction models for a variety of data likelihoods. I discuss results on text and image datasets, as well as neural spike trains. Next, I present a multi-view framework that combines neural networks and probabilistic CCA to estimate shared and view-specific latent structure of paired samples of histological images and gene expression levels. The model is trained end-to-end to estimate all parameters simultaneously, and I show that the latent variables capture interpretable hidden structure, such as tissue-specific and morphological variation. Finally, I discuss lowering the computational cost of online Bayesian filtering of time series with abrupt changes in structure, called changepoints. In time series with multiple data sources and associated costs, one can trade the cost of collecting an observation against the quality or "fidelity" of the measurement and how this fidelity affects the estimation of changepoints. Our framework makes cost-sensitive decisions about which data fidelity to use based on maximizing information gain with respect to the posterior distribution over changepoints. I present results on streaming video and audio data. I conclude by discussing several lines of work extending these projects.

1 0

CS Colloquium Speaker: Ellen Vitercik on Monday, Feb 22
by Emily C. Lawrence 16 Feb '21

16 Feb '21

CS Colloquium Speaker Ellen Vitercik, Carnegie Mellon University Host: Sanjeev Arora Date: Monday, February 22 Time: 12:30pm EST Event page: [ https://www.cs.princeton.edu/events/26042 | https://www.cs.princeton.edu/events/26042 ] Register for the Zoom webinar here: [ https://princeton.zoom.us/webinar/register/WN_gbMN0OUJQRainEiRTS4VYQ | https://princeton.zoom.us/webinar/register/WN_gbMN0OUJQRainEiRTS4VYQ ] Title: Integrating Machine Learning into Algorithm Design Abstract: An important property of those algorithms that are typically used in practice is broad applicability—the ability to solve problems across diverse domains. However, the default, out-of-the-box performance of these algorithms can be unsatisfactory, with slow runtime, poor solution quality, and even negative long-term social ramifications. In practice, there is often ample data available about the types of problems that an algorithm will be run on, data that can potentially be harnessed to fine-tune the algorithm’s performance. We therefore need principled approaches for using this data to obtain strong application-specific performance guarantees. In this talk, I will give an overview of my research that provides practical methods built on firm theoretical foundations for incorporating machine learning and optimization into the process of algorithm design, selection, and configuration. I will describe my contributions across several diverse domains, including integer programming, clustering, mechanism design, and computational biology. As I will demonstrate, these seemingly disparate areas are connected by overarching structure which implies broadly-applicable guarantees. Bio: Ellen Vitercik is a PhD student at Carnegie Mellon University where she is co-advised by Maria-Florina Balcan and Tuomas Sandholm. Her research revolves around artificial intelligence, algorithm design, and the interface between economics and computation, with a particular focus on machine learning theory. Among other honors, she is a recipient of the Exemplary Artificial Intelligence Track Paper Award at EC'19, the Best Presentation by a Student or Postdoctoral Researcher Award at EC'19, the NSF Graduate Research Fellowship, the IBM PhD Fellowship, the Fellowship in Digital Health from CMU's Center for Machine Learning and Health, and the Teaching Assistant of the Year Award from CMU's Machine Learning Department. This talk will be recorded.

1 0

2/16/2021 ECE Seminar - Yuan Tian, University of Virginia
by Emily C. Lawrence 15 Feb '21

15 Feb '21

ECE SEMINAR Speaker: Yuan Tian, University of Virginia Title: Protecting User Security and Privacy in Emerging Platforms Day: Tuesday, February 16, 2021 Time: 4:30 pm Room: Please register at [ https://princeton.zoom.us/meeting/register/tJMod-CvrjIoGt3_Mbu5LZwjr59TagXs… | https://princeton.zoom.us/meeting/register/tJMod-CvrjIoGt3_Mbu5LZwjr59TagXs… ] Host: Niraj Jha Abstract: The evolution of apps on new platforms such as mobile, web, and the Internet of Things (IoT) are bringing more functionality and convenience to people; however, these new platforms also expose users to security and privacy risks. For example, mobile and IoT devices use sensors and machine learning to provide richer functionality, but these features may violate users’ security and privacy. Researchers and developers are spending much effort to protect the users, but unauthorized information leakage is still rampant, especially when new features or new techniques are introduced. To resolve these problems, I work on changing the way in which platform designers think about designing secure systems, assisting the developers about secure system implementations, and creating technologies to facilitate better security decision-making. In this talk, I’ll present my example projects in two thrusts: (1) identifying and understanding new threats and (2) designing and implementing secure and privacy-preserving systems. In the first thrust, I will use voice-controlled devices as an example to show how we identify new security and privacy threats on devices powered by machine learning. Our solutions have been adopted by the device vendors such as Google and Amazon. In the second thrust, I introduce our efforts in building secure and privacy-preserving systems for the Internet of Things. I performed program analysis to discover vulnerabilities of current permission systems in third-party IoT apps. With insights from program analysis and natural language processing, I proposed principles and implemented a privacy-preserving system to share the least privilege information to third-party apps without affecting their functionality. In general, I hope to bring low-level privacy enhancements to the users through thorough design, efficient implementation, and usable interfaces. Bio: Yuan Tian is an Assistant Professor of Computer Science at the University of Virginia. Before joining UVA, she obtained her Ph.D. from Carnegie Mellon University in 2017 and interned at Microsoft Research, Facebook, and Samsung Research. Her research interests involve security and privacy and its interactions with computer systems, machine learning, and human-computer interaction. Her current research focuses on developing new technologies for protecting user privacy, particularly in the areas of mobile systems and the Internet of Things. Her work has generated real-world impact as countermeasures and design changes have been integrated into platforms (such as Android, Chrome, SmartThings, Azure, and iOS), and also impacted the security recommendations of standard organizations such as Internet Engineering Task Force (IETF) and World Wide Web Consortium (W3C). She is a recipient of NSF CAREER award 2020, NSF CRII award 2019, Amazon AI Faculty Fellowship 2019, CSAW Best Security Paper Award 2019, Rising Stars in EECS 2016, and Black Hat Future Female Leaders in Cyber Security 2015. Her research has appeared in top-tier venues in Security, Machine Learning, and System. Her projects have been covered by media outlets such as IEEE Spectrum, Forbes, Fortune, Wired, and Telegraph.

1 0

CS Distinguished Colloquium Speaker: Christos Papadimitriou, Friday, Feb 19 at 12:30
by Emily C. Lawrence 12 Feb '21

12 Feb '21

CS Distinguished Colloquium Speaker Christos Papadimitriou, Columbia University Host: Sanjeev Arora Date: Friday, February 19 Time: 12:30pm EST Event page: [ https://www.cs.princeton.edu/events/26038 | https://www.cs.princeton.edu/events/26038 ] Register for Zoom webinar here: [ https://princeton.zoom.us/webinar/register/WN_oQAMcljNQbWYTJcaDmTKDQ | https://princeton.zoom.us/webinar/register/WN_oQAMcljNQbWYTJcaDmTKDQ ] Title: Language, Brain, and Computation Abstract: How does the brain beget the mind? How do molecules, cells and synapses effect reasoning, intelligence, language? Despite dazzling progress in experimental neuroscience, as well as in cognitive science at the other extreme of scale, we do not seem to be making progress in the overarching question -- the gap is huge and a completely new approach seems to be required. As Richard Axel recently put it: "We don't have a logic for the transformation of neural activity into thought [...]." What kind of formal system would qualify as this "logic"? I will introduce the Assembly Calculus, a computational system whose basic data structure is the assembly -- assemblies are large populations of neurons representing concepts, words, ideas, episodes, etc., The Assembly Calculus is biologically plausible in the following two orthogonal senses: Its primitives are properties of assemblies observed in experiments or useful for explaining experiments, and can be provably (through both mathematical proof and simulations) "compiled down" to the activity of neurons and synapses. I will also explain why the Assembly Calculus is especially powerful in exploring how language happens in the brain. Bio: Christos Harilaos Papadimitriou is the Donovan Family Professor of Computer Science at Columbia University. Before joining Columbia in 2017, he was a professor at UC Berkeley for the previous 22 years, and before that he taught at Harvard, MIT, NTU Athens, Stanford, and UCSD. He has written five textbooks and many articles on algorithms and complexity, and their applications to optimization, databases, control, AI, robotics, economics and game theory, the Internet, evolution, and the brain. He holds a PhD from Princeton (1976), and eight honorary doctorates, including from ETH, University of Athens, EPFL, and Univ. de Paris Dauphine. He is a member of the National Academy of Sciences of the US, the American Academy of Arts and Sciences, and the National Academy of Engineering, and he has received the Knuth prize, the Go"del prize, the von Neumann medal, as well as the 2018 Harvey Prize by Technion. In 2015 the president of the Hellenic republic named him commander of the order of the Phoenix. He has also written three novels: “Turing ,” “Logicomix” and his latest “Independence.”

1 0

Alex Upjohn Beatson will present his Pre-FPO "Learned surrogates and stochastic gradients for accelerating engineering modeling, simulation, and design" on Monday, February 8, 2021 at 11am via Zoom.
by Nicki Mahler 05 Feb '21

05 Feb '21

Alex Upjohn Beatson will present his Pre-FPO "Learned surrogates and stochastic gradients for accelerating engineering modeling, simulation, and design" on Monday, February 8, 2021 at 11am via Zoom. Zoom Link: [ https://princeton.zoom.us/j/99264611511 | https://princeton.zoom.us/j/99264611511 ] The members of his committee are as follows: Readers - Szymon Rusinkiewicz, Sigrid Adriaenssens (Civil and Environmental Engineering) Examiners - Ryan Adams (advisor), Elad Hazan, Mark Girolami (Cambridge University) Title : Learned surrogates and stochastic gradients for accelerating engineering modeling, simulation, and design Abstract : Numerical methods, such as discretization-based methods for solving ODEs and PDEs, allow us to model and design complex devices, structures, and systems. However, this is often very costly in terms of both computation and the time of the expert who must specify the physical model, the discretization, the solver, etc. In this talk I will present a line of work using deep learning and stochastic gradient estimation to speed up solution and optimization of systems characterized by such numerical methods. First, I will present composable energy surrogates, in which neural surrogates are trained to model a potential energy in sub-components or sub-domains of a PDE, and then composed or stitched together to solve a larger system by minimizing the sum of potentials across components. This allows surrogate modeling without costly ground-truth simulation of the full system, as training data are generated by performing finite element analysis with individual components. We show that these surrogates can accelerate simulation of parametric meta-materials and produce accurate macroscopic behavior when composed. Next, I will discuss randomized telescoping gradient estimators, which provide unbiased gradient estimators for objectives which are the limit of a sequence of increasingly accurate, increasingly costly approximations. These estimators represent the limit as a telescoping sum and sample linear combinations of terms to provide cheap unbiased estimates. We discuss conditions which permit finite variance and computation, optimality of certain estimators within this class, and application to problems in numerical modeling and machine learning. Lastly, I will discuss how these lines of work can be extended to reduce not just the computational burden but also the burden on the engineer carrying out modeling and design.

1 0

Ethan Tseng will present his General Exam "Perceptually Motivated Camera Design" on February 5, 2021 at 12pm via Zoom.
by Nicki Mahler 01 Feb '21

01 Feb '21

Ethan Tseng will present his General Exam "Perceptually Motivated Camera Design" on February 5, 2021 at 12pm via Zoom. Zoom link: [ https://princeton.zoom.us/j/97997704193 | https://princeton.zoom.us/j/97997704193 ] The members of Ethan's committee are as follows: Felix Heide (adviser), Szymon Rusinkiewicz, Adam Finkelstein Abstract: Machine vision relies on images captured by low-level optics and hardware, however, image acquisition and image processing are often considered as disjoint tasks to be solved in isolation. Conventional cameras are designed to capture the most visually pleasing images to human consumers without concern for the post-processing algorithms, and similarly, computer vision algorithms are often expected to perform equally well regardless of the camera used. In this talk, I present my recent research that aims to bridge this gap by learning the imaging stack in an end-to-end fashion. First, I demonstrate this framework by designing a diffractive optic imager for capturing high-dynamic range pictures in a single shot. Next, I introduce an end-to-end camera design system that jointly optimizes compound lenses, image signal processors (ISP), and downstream post-processors with respect to high-level computer vision loss functions. Finally, I present a neural nano-optic imager that captures high-quality RGB images on par with commercial compound lenses but is over 500000 times smaller in size. This state-of-the-art performance was achieved through end-to-end learning of the nano-optic's physical structure in conjunction with a novel, neural feature-based image reconstruction algorithm. Link to Abstract and Reading List: [ https://docs.google.com/document/d/1h-t_yQtzCmrmEA0IqiUym0VJbtYgAO__3F1Ho6L… | https://docs.google.com/document/d/1h-t_yQtzCmrmEA0IqiUym0VJbtYgAO__3F1Ho6L… ] 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

Ben Burgess will present his Pre FPO "Reverse Engineering Real World Trusted Systems" on Thursday, January 28, 2021 at 4pm via Zoom.
by Nicki Mahler 25 Jan '21

25 Jan '21

Ben Burgess will present his Pre FPO " Reverse Engineering Real World Trusted Systems" on Thursday, January 28, 2021 at 4pm via Zoom. Zoom Information: [ https://princeton.zoom.us/j/98643121461 | https://princeton.zoom.us/j/98643121461 ] Meeting ID: 986 4312 1461 Examiners: Ed Felten (adviser), Arvind Narayanan, Nick Feamster (University of Chicago) Readers: Ed Felten (adviser), Prateek Mittal, Danny Huang (New York University) All are welcome to attend. Abstract: Overview In this talk, I will present two studies that focus on evaluating the security and privacy guarantees of real world systems that have the potential to affect large groups of people. The first study I will discuss evaluates the security of trunking radio systems which are used as the primary communication system for almost all public safety agencies. These radio systems provide higher availability guarantees than cellular phones or conventional two way radios while also promising completely private communications. In this talk, I will discuss the vulnerabilities these systems suffer either due to weak encryption ciphers or flawed protocol design which can allow an adversary to recover sensitive voice or data traffic. The second study I will present looks at the various exam proctoring software solutions that institutions are contracting to proctor their exams during COVID. I will provide a detailed analysis of the impact of these solutions on student privacy along with a look at the security guarantees they provide the institution. Police Radio For this part of the talk, I will focus on trunking radio systems designed by Motorola and Harris as they are the largest vendors in the public safety communication market. To protect voice communications, Motorola offers several different encryption options for the user to choose from including common ciphers such as DES and AES along with proprietary options such as their Advanced Digital Privacy (ADP) cipher. An earlier study by Glass et al. reverse engineered ADP and demonstrated that ADP is simply a wrapper around RC4 which has many known vulnerabilities. We demonstrate that the underlying voice traffic can be easily recovered from real world deployments using low cost digital TV receivers despite multipath interference challenges. We then demonstrate that both ADP and DES encrypted traffic can be recovered from these captures at a low cost using either desktop GPU hardware or rentable FPGA clusters. Finally, we look at the over the air rekey (OTAR) mechanism Motorola implements which allows the user to rekey certain voice channels when a radio is lost or stolen. We find serious issues with the OTAR protocol which allows a user to easily compromise this mechanism once they are able to decrypt voice traffic. By compromising the OTAR protocol, an adversary would be able to maintain persistent access to the encrypted communications on the system with no further work until all of the radios were recalled and manually rekeyed using a hardware loader. We extend this study to look at the other major offering in the public safety market, OpenSky by Harris. This offering has seen widespread adoption across the public safety market due to its perceived security despite lacking interoperability with the Motorola standard. We reverse engineer the protocol specification of the Harris solution to provide the first publicly available look at the underlying design. We demonstrate that traffic is not always encrypted as is widely believed by the agencies deploying the system. Additionally, we design an attack against this system which allows for an adversary to both intercept and forge traffic between the subscriber unit and base station. Exam Proctoring Software Study For this part of the talk, I will provide a detailed analysis of the impact of the major exam proctoring software offerings on student privacy along with a look at the security guarantees they provide. I will focus primarily on one of the most popular offerings, Examplify, which is being used to proctor the bar exam in many states. I will discuss several different exploits for it which allow the user to run the software in a virtual machine without detection, extract the exam contents without being logged, and extract the answers for the exam in certain cases. I will then discuss the details of the facial recognition models these solutions use to detect cheating and the potential racial biases of these models. Finally, I will discuss the impact on student privacy by looking at what the system service logs and shares with the developer.

1 0