Select a calendar:
Filter February Events by Event Type:
SUNMONTUEWEDTHUFRISAT
Events for February
-
Theory Lunch
Thu, Feb 06, 2020 @ 12:15 PM - 02:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Kobbi Nissim, Georgetown University
Talk Title: Legal Theorems of Privacy
Abstract: There are significant gaps between legal and technical thinking around data privacy. Technical standards such as k-anonymity and differential privacy are described using mathematical language and strive for mathematical rigor whereas legal standards are not rigorous from a mathematical point of view and often resort to concepts such as de-identification and anonymization which they only partially define. As a result, arguments about the adequacy of technical privacy measures for satisfying legal privacy often lack rigor, and their conclusions are uncertain. The uncertainty is exacerbated by a litany of successful privacy attacks on privacy measures thought to meet legal expectations but then shown to fall short of doing so.
We ask whether it is possible to introduce mathematical rigor into such analyses so as to make formal claims and prove "legal theorems" that technical privacy measures meet legal expectations. For that, we explore some of the gaps between these two very different approaches, and present initial strategies towards bridging these gaps. In particular, we focus on the concept of singling out from the EU's General Data Protection Regulation (GDPR). To capture this concept, we define a new type of privacy attack, predicate singling out, where an adversary finds a predicate matching exactly one row in a database with probability significantly better then a statistical baseline. We then argue that any data release mechanism that purports to "render anonymous" data under the GDPR should prevent predicate singling out. Hence, the concept has legal consequences as it can be used as a yardstick for arguing whether data release mechanisms meet the GDPR standard of data anonymization.
Biography: Professor Kobbi Nissim is a McDevitt Chair at the department of Computer Science, Georgetown University and affiliated with Georgetown Law. Nissim's work is focused on the mathematical formulation and understanding of privacy. His work from 2003 and 2004 with Dinur and Dwork initiated rigorous foundational research of privacy and in 2006 he introduced differential privacy with Dwork, McSherry and Smith. Nissim was awarded the Caspar Bowden Privacy for research in Privacy Enhancing Technology in 2019, the Gödel Prize in 2017, IACR TCC Test of Time Awards in 2016 and in 2018, and the ACM PODS Alberto O. Mendelzon Test-of-Time Award in 2013.
Host: Shaddin Dughmi
Location: 213
Audiences: Everyone Is Invited
Contact: Cherie Carter
-
CS Distinguished Lecture: Karon MacLean (University of British Columbia) - Making Haptics and its Design Accessible
Thu, Feb 06, 2020 @ 04:00 PM - 05:20 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Karon MacLean, University of British Columbia, Canada
Talk Title: Making Haptics and its Design Accessible
Series: Computer Science Distinguished Lecture Series
Abstract: Today's advances in tactile sensing and wearable, IOT and context-aware computing are spurring new ideas about how to configure touch-centered interactions in terms of roles and utility, which in turn expose new technical and social design questions. But while haptic actuation, sensing and control are improving, the difficulties of incorporating them into a real-world design process poses a major obstacle to adoption in everyday technology.
In this talk I'll overview highlights chosen from of an ongoing effort to understand how to support haptic designers and end-users. These include online experimental design tools, DIY open sourced hardware and accessible means of creating, for example, expressive physical robot motions and evolve physically sensed expressive tactile languages, and major community-based studies of design practice.
To accelerate design practice, we put our systems, designs and datasets online. A central and evolving piece of our larger openhaptics effort is Haptipedia, an expert-sourced, community-based browsable visualization of historical haptic inventions as a resource to future designers.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
Biography: Karon MacLean is Professor in Computer Science at UBC, with degrees in Biology and Mechanical Engineering (BSc, Stanford; M.Sc. / Ph.D, MIT) and time spent as a professional robotics engineer (Center for Engineering Design, University of Utah) and haptics / interaction researcher (Interval Research, Palo Alto). At UBC since 2000, MacLean's research specializes in haptic (touch) interaction: cognitive, sensory and affective design for people interacting with the computation we touch, emote and move with and learn from, from robots to handheld devices and the situated environment. MacLean leads UBC's Designing for People interdisciplinary research cluster and CREATE graduate training program (25 researchers spanning 11 departments and 5 faculties - dfp.ubc.ca), is Special Advisor, Innovation and Knowledge Mobilization to UBC's Faculty of Science, and will co-chair ACM UIST (User Interface Software and Technology) in 2020.
Host: Heather Culbertson
Location: Henry Salvatori Computer Science Center (SAL) - 101
Audiences: Everyone Is Invited
Contact: Computer Science Department
-
Computer Science General Faculty Meeting
Wed, Feb 12, 2020 @ 12:00 PM - 02:00 PM
Thomas Lord Department of Computer Science
Receptions & Special Events
Bi-Weekly regular faculty meeting for invited full-time Computer Science faculty only. Event details emailed directly to attendees.
Location: Ronald Tutor Hall of Engineering (RTH) - 526
Audiences: Invited Faculty Only
Contact: Assistant to CS chair
-
Theory Lunch
Thu, Feb 13, 2020 @ 12:15 PM - 02:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Salil Vadhan, Harvard University
Talk Title: Derandomization Beyond Connectivity: High-Precision Estimation of Random Walks and Laplacian Solvers in Small Space
Abstract: I will describe a series of works that attacks the derandomization of space-bounded computation (e.g. seeking to prove RL=L) using a combination of ideas from the literature on time-efficient Laplacian solvers (Spielman and Teng, STOC '04; Peng and Spielman, STOC '14; Cheng et al. '15; Cohen et al. FOCS '16, STOC '17, FOCS '18) with ones used to show that Undirected S-T Connectivity is in deterministic logspace (Reingold, STOC '05 and JACM '08; Rozenman and Vadhan, RANDOM '05).
In particular, we obtain deterministic, nearly logarithmic-space algorithms for (a) estimating random walk probabilities to within polynomially small error and (b) approximately solving linear systems given by graph Laplacians, with both results holding for Eulerian directed graphs and hence also undirected graphs. Previously both of these problems were known to be solvable for general directed graphs by randomized algorithms in logarithmic space (Aleliunas et al. FOCS '79; Doron, Le Gall, and Ta-Shma RANDOM '17), and hence by deterministic algorithms using space O(log^{3/2} N) (Saks and Zhou, FOCS '95 and JCSS '99).
Joint works with Murtagh, Reingold, and Sidford (FOCS '17 and RANDOM '19) and Ahmadinejad, Kelner, Murtagh, Peebles, and Sidford (arXiv:1912.04524)
Biography: Salil Vadhan is Vicky Joseph Professor of Computer Science and Applied Mathematics at Harvard University. After completing his undergraduate degree in Mathematics and Computer Science at Harvard in 1995, he obtained his PhD in Applied Mathematics from Massachusetts Institute of Technology in 1999, where his advisor was Shafi Goldwasser. His research centers around the interface between computational complexity theory and cryptography. He focuses on the topics of pseudorandomness and zero-knowledge proofs. His work on zig-zag product, with Omer Reingold and Avi Wigderson, was awarded the 2009 Gödel Prize.
Host: Shaddin Dughmi
Location: Henry Salvatori Computer Science Center (SAL) - 213
Audiences: Everyone Is Invited
Contact: Cherie Carter
-
CS Colloquium: Scott Niekum (UT Austin) - Scaling Probabilistically Safe Learning to Robotics
Fri, Feb 14, 2020 @ 02:00 PM - 03:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Scott Niekum, The University of Texas at Austin
Talk Title: Scaling Probabilistically Safe Learning to Robotics
Series: Computer Science Colloquium
Abstract: Before learning robots can be deployed in the real world, it is critical that probabilistic guarantees can be made about the safety and performance of such systems. In recent years, safe reinforcement learning algorithms have enjoyed success in application areas with high-quality models and plentiful data, but robotics remains a challenging domain for scaling up such approaches. Furthermore, very little work has been done on the even more difficult problem of safe imitation learning, in which the demonstrator's reward function is not known. This talk focuses on new developments in three key areas for scaling safe learning to robotics: (1) a theory of safe imitation learning; (2) scalable reward inference in the absence of models; (3) efficient off-policy policy evaluation. The proposed algorithms offer a blend of safety and practicality, making a significant step towards safe robot learning with modest amounts of real-world data.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
Biography: Scott Niekum is an Assistant Professor and the director of the Personal Autonomous Robotics Lab (PeARL) in the Department of Computer Science at UT Austin. He is also a core faculty member in the interdepartmental robotics group at UT. Prior to joining UT Austin, Scott was a postdoctoral research fellow at the Carnegie Mellon Robotics Institute and received his Ph.D. from the Department of Computer Science at the University of Massachusetts Amherst. His research interests include imitation learning, reinforcement learning, and robotic manipulation. Scott is a recipient of the 2018 NSF CAREER Award and 2019 AFOSR Young Investigator Award.
Host: Stefanos Nikolaidis
Location: Henry Salvatori Computer Science Center (SAL) - 101
Audiences: Everyone Is Invited
Contact: Computer Science Department
-
CS Colloquium: Cheng Tan (Courant Institute / New York University) - Auditing Outsourced Services
Tue, Feb 18, 2020 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Cheng Tan, Courant Institute / New York University
Talk Title: Auditing Outsourced Services
Series: CS Colloquium
Abstract: How can users of a cloud service verify that the service truly performs as promised? This question is vital today because clouds are complicated black boxes, running in different administrative domains from users. Their correctness can be undermined by internal corruptions---misconfigurations, operational mistakes, insider attacks, unexpected failures, or adversarial control at any layer of the execution stack.
This talk will present verifiable infrastructure, a framework that lets users audit outsourced applications and services. I will introduce two systems: Orochi and Cobra, which verify the execution of, respectively, untrusted servers and black-box databases. Orochi and Cobra introduce various techniques, including deduplicated re-execution, consistent ordering verification, GPU accelerated pruning, and others. Beyond these two systems, I will also discuss verifiable infrastructure more generally.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
Biography: Cheng Tan is a computer science Ph.D. candidate in the Courant Institute at New York University. His interests are in operating systems, networked systems, and security. His work on the Efficient Server Audit Problem was awarded best paper at SOSP 2017. His work on data center network troubleshooting at Microsoft Research has been deployed globally in more than 30 data centers in Microsoft Azure.
Host: Barath Raghavan
Location: Olin Hall of Engineering (OHE) - 132
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium: Jiapeng Zhang (Harvard) - Sunflowers and Their Applications in Computer Science and Mathematics
Thu, Feb 20, 2020 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Jiapeng Zhang, Harvard University
Talk Title: Sunflowers and Their Applications in Computer Science and Mathematics
Series: CS Colloquium
Abstract: The sunflower is a simple notion in combinatorics, originally invented and studied by Erdos and Rado in 1960. Surprisingly, it has deep connections to fundamental problems in computer science, such as matrix multiplication, efficient data structures, computational complexity and cryptography. In my talk, I will explain our new results on sunflowers, how ideas emerging from computer science were critical in the proof, and how our new techniques can help shed light on some central problems in computer science and mathematics.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Jiapeng Zhang is a postdoc at Harvard with Prof. Salil Vadhan. He did his PhD at UC San Diego with Prof. Shachar Lovett. His research focuses on boolean function analysis, computational complexity, learning theory and cryptography.
Host: Shaddin Dughmi
Location: Olin Hall of Engineering (OHE) - 132
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
MASCLE Machine Learning Seminar: Rose Yu (Northeastern University) - Physics Guided AI for Learning Spatiotemporal Dynamics
Thu, Feb 20, 2020 @ 04:00 PM - 05:20 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Rose Yu, Northeastern University
Talk Title: Physics Guided AI for Learning Spatiotemporal Dynamics
Series: Machine Learning Seminar Series hosted by USC Machine Learning Center
Abstract: Applications such as sports, climate science, and aerospace engineering require learning complex dynamics from large-scale spatiotemporal data. Such data is often non-linear, non-Euclidean, high-dimensional, and demonstrates complicated dependencies. Existing machine learning frameworks are still insufficient to learn spatiotemporal dynamics as they often fail to exploit the underlying physics principles. I will demonstrate how to inject physical knowledge in AI to deal with challenges such as non-linear dynamics, non-Euclidean geometry, and multi-resolution structure. I will showcase the application of these methods to problems such as accelerating turbulence simulations, imitating basketball gameplay and combating ground effect in quadcopter landing.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
Biography: Dr. Yu is an Assistant Professor in the Khoury College of Computer Sciences at Northeastern University. Previously, she was a postdoctoral researcher at Caltech Computing and Mathematical Sciences. She earned her Ph.D. in Computer Sciences at the University of Southern California. Her research focuses on advancing machine learning techniques for large-scale spatiotemporal data, with a particular emphasis on physics-guided AI. Among her awards, she has won Google Faculty Research Award, the NSF CRII award, best dissertation award in USC, best paper award at the NeurIPS time series workshop, and was nominated as one of the 'MIT Rising Stars in EECS'.
Host: Yan Liu
Location: Henry Salvatori Computer Science Center (SAL) - 101
Audiences: Everyone Is Invited
Contact: Computer Science Department
-
CS Colloquium: Michael Everett (MIT) - Fully Autonomous Robot Navigation in Human Environments
Mon, Feb 24, 2020 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Michael Everett, MIT
Talk Title: Fully Autonomous Robot Navigation in Human Environments
Series: CS Colloquium
Abstract: Today's robots are still quite limited in their ability to process information about multiple other objects in order to plan safe and efficient motions through previously unseen environments. Major technical challenges are currently sidestepped by restrictive engineering solutions (e.g., preventing humans from working alongside factory robots, collecting detailed prior maps in every intended operating environment). This talk will present frameworks that enable long-term autonomy for robots embedded among pedestrians and context-guided exploration in new environments. Furthermore, it will discuss future research directions toward safely training and deploying robots in our society.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Michael Everett is a final-year PhD Candidate at MIT working with Prof. Jonathan How. He received the SM degree (2017) and the SB degree (2015) from MIT in Mechanical Engineering. His research addresses fundamental gaps in the connection of machine learning and real mobile robotics, with recent emphasis on developing the theory of safety/robustness of learned modules. His works have won the Best Paper Award on Cognitive Robotics at IROS 2019, the Best Student Paper Award and finalist for the Best Paper Award on Cognitive Robotics at IROS 2017, and finalist for the Best Multi-Robot Systems Paper Award at ICRA 2017. He has been interviewed live on the air by BBC Radio and his robots were featured by Today Show, Reuters, and the Boston Globe.
Host: Nora Ayanian
Location: Ronald Tutor Hall of Engineering (RTH) - 109
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium: Robin Jia (Stanford University) - Building Robust Natural Language Processing Systems
Tue, Feb 25, 2020 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Robin Jia, Stanford University
Talk Title: Building Robust Natural Language Processing Systems
Series: CS Colloquium
Abstract: While modern NLP systems have achieved outstanding performance on static benchmarks, they often fail catastrophically when presented with inputs from different sources or inputs that have been adversarial perturbed. This lack of robustness exposes troubling gaps in current models' understanding capabilities, and poses challenges for deployment of NLP systems in high-stakes situations. In this talk, I will demonstrate that building robust NLP systems requires reexamining all aspects of the current model building paradigm. First, I will show that adversarially constructed test data reveals vulnerabilities that are left unexposed by standard evaluation methods. Second, I will demonstrate that active learning, in which data is adaptively collected based on a model's current predictions, can significantly improve the ability of models to generalize robustly, compared to the use of static training datasets. Finally, I will show how to train NLP models to produce certificates of robustness---guarantees that for a given example and combinatorially large class of possible perturbations, no perturbation can cause a misclassification.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Robin Jia is a sixth-year Ph.D. student at Stanford University advised by Percy Liang. His research interests lie broadly in building natural language processing systems that can generalize to unexpected test-time inputs. Robin's work has received an Outstanding Paper Award at EMNLP 2017 and a Best Short Paper Award at ACL 2018. He has been supported by an NSF Graduate Research Fellowship.
Host: Xiang Ren
Location: Olin Hall of Engineering (OHE) - 132
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
Computer Science General Faculty Meeting
Wed, Feb 26, 2020 @ 12:00 PM - 02:00 PM
Thomas Lord Department of Computer Science
Receptions & Special Events
Bi-Weekly regular faculty meeting for invited full-time Computer Science faculty only. Event details emailed directly to attendees.
Location: Ronald Tutor Hall of Engineering (RTH) - 526
Audiences: Invited Faculty Only
Contact: Assistant to CS chair
-
CS Colloquium: Minjoon Seo (University of Washington) - Web-Scale Neural Memory towards Universal Knowledge Interface
Thu, Feb 27, 2020 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Minjoon Seo, University of Washington
Talk Title: Web-scale Neural Memory towards Universal Knowledge Interface
Series: CS Colloquium
Abstract: Modern natural language tasks are increasingly dependent on external world knowledge. My PhD study has particularly focused on three challenges in this literature: handling unstructured knowledge, being scalable, and reasoning over knowledge data. I will mainly discuss my recent and on-going work on a web-scale neural memory that tackles all of the three challenges, and show how it serves as an effective interface for interacting with the world knowledge. I will conclude with an argument that designing a seamless and universal knowledge interface is a crucial research goal that can better address knowledge-dependency problem in machine learning tasks.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Minjoon Seo is a final-year Ph.D. student in the Allen School of Computer Science & Engineering at the University of Washington, advised by Hannaneh Hajishirzi and Ali Farhadi. His research interest has been mostly in the learning model for the extraction of (IE), the access to (QA), and the interplay of (Reasoning) knowledge in various forms of language data. He is supported by Facebook Fellowship and AI2 Key Scientific Challenges Award. He co-organizes the Workshop on Machine Reading for Question Answering (MRQA) and the Workshop on Representation Learning for NLP (RepL4NLP).
Host: Xiang Ren
Location: Olin Hall of Engineering (OHE) - 132
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
PhD Defense - Ayush Jaiswal
Fri, Feb 28, 2020 @ 01:30 PM - 03:30 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Candidate: Ayush Jaiswal
Date: Friday, February 28, 2020
Time: 1:30 PM - 3:30 PM
Location: SAL 213
Committee: Premkumar Natarajan (Chair), Ram Nevatia, Cauligi S. Raghavendra
Title: Invariant Representation Learning for Robust and Fair Predictions
Abstract:
Learning representations that are invariant to nuisance factors of data improves robustness of machine learning models, and promotes fairness for factors that represent biasing information. This view of invariance has been adopted for deep neural networks (DNNs) recently as they learn latent representations of data by design. Numerous methods for invariant representation learning for DNNs have emerged in recent literature, but the research problem remains challenging to solve: existing methods achieve partial invariance or fall short of optimal performance on the prediction tasks that the DNNs need to be trained for.
This thesis presents novel approaches for inducing invariant representations in DNNs by effectively separating predictive factors of data from undesired nuisances and biases. The presented methods improve the predictive performance and the fairness of DNNs through increased invariance to undesired factors. Empirical evaluation on a diverse collection of benchmark datasets shows that the presented methods achieve state-of-the-art performance.
Application of the invariance methods to real-world problems is also presented, demonstrating their practical utility. Specifically, the presented methods improve nuisance-robustness in presentation attack detection and automated speech recognition, fairness in face-based analytics, and generalization in low-data and semi-supervised learning settings.
Location: Henry Salvatori Computer Science Center (SAL) - 213
Audiences: Everyone Is Invited
Contact: Lizsl De Leon
