(The talk will be given in English)

Speaker:     Dr. Yair Carmon
                     Stanford University

Monday, December 23^rd, 2019
15:00 - 16:00

Room 011, Kitot Bldg., Faculty of Engineering

Fundamental limits of modern machine learning and how to get around them

Abstract

The computational barriers concern nonconvex optimization: we prove unconditional lower bounds on the oracle complexity of finding stationary points using (stochastic) gradient methods. To bypass these lower bounds, we develop a “convex until proven guilty” principle that leverages additional smoothness available for many problems. Combining it with classical momentum techniques, we obtain an accelerated method for nonconvex optimization.

The statistical barriers concern the sample complexity of adversarially robust learning. We prove that unlabeled data allows us to circumvent an information theoretic gap between robust and standard classification in a Gaussian model. Our analysis directly leads to a general robust self-training procedure; we use it to improve state-of-the-art performance on the challenging and extensively studied CIFAR-10 adversarial robustness benchmark by up to 7 points.

Short Bio
Yair Carmon is a PhD student at Stanford University working with Prof. John Duchi and Prof. Aaron Sidford. His research focuses on the foundations of machine learning and draws on optimization, algorithm analysis, statistics and information theory. Yair received an M.Sc. from the Technion in 2015, under the supervision of Prof. Shlomo Shamai and Prof. Tsachy Weissman. Between 2009 and 2015 he held several positions as an algorithm engineer and research team leader, working on communications, signal processing and computer vision. His awards include the Stanford Graduate Fellowship, the Numerical Technologies Fellowship, and membership in the Technion Excellence Program.   ‏

(The talk will be given in English)

Speaker:     Dr. Wasim Huleihel
                    Department of Electrical Engineering, Tel Aviv University

SUNDAY, January 5^th, 2020
15:00 - 16:00

Room 011, Kitot Bldg., Faculty of Engineering

Statistical and Computational Tradeoffs in Interactive Clustering

Abstract

The ever-increasing size and complexity of modern data sets pose many challenges for statistical inference, while classical statistical analysis has lagged behind. Among these challenges are fundamental questions related to the interplay and tradeoffs between statistical and computational requirements, which emerge due to the presence of complex underlying combinatorial structures in modern high-dimensional problems. In this talk, I will introduce a semi-supervised formulation of the clustering problem that allow active querying during data segmentation. Clever implementation of interactive querying framework can improve the accuracy of clustering and help in inferring labels of large amount of data by issuing only a small number of queries. For this problem, I will characterize the statistical and computational boundaries on the minimal number of pairwise queries needed to recover (possibly overlapping) latent clusters, under both adversarial and statistical modeling assumptions.

Short Bio
Wasim Huleihel received the B.Sc. and M.Sc. degrees in 2012 and 2013 from Ben-Gurion University, respectively, and received his PhD degree in 2017 from the Technion, all in electrical engineering. He spent two years as a postdoctoral fellow in the Research Laboratory of Electronics (RLE) at the Massachusetts Institute of Technology (MIT), and is currently a postdoctoral fellow in the Department of Electrical Engineering - Systems at Tel-Aviv university. His research interests include high dimensional statistics, theoretical machine learning, information theory, and statistical signal processing.
Wasim is the recipient of the MIT - Technion Postdoctoral Fellowship, Viterbi fellowship, the Vatat postdoctoral scholarship, Vatat fellowship for excellent PhD students, the Advanced Communication Center (ACC) Feder Family Award for outstanding research work in the field of communication technologies (first prize), coauthor of the Best Student Paper Award in the 31rd Annual Conference on Learning Theory (COLT'18), B.Sc. and M.Sc. graduations with honor, M.Sc. Rector's List Award, and Technion excellent tutor awards.‏

מהנדס תוכנה, פיזיקאית, אלגוריתמאי, מהנדס חומרה, מהנדסת אפליקציה ומפתח מערכות מידע נכנסים לבר...חברת Orbotech מזמינים אתכם, סטודנטים מכל מקצועות ההנדסה, למסע אל עתיד הטכנולוגיה ומעבר לה!

6 דוברים, 6 נושאים טכנולוגים שונים, כ-12 דק' כל הרצאה, ערב חד פעמי.

מיקום: StanliJoys (בכניסה למעונות ברושים, ג'ורג' וייז 20)

תאריך: יום שלישי ה-17/12 החל מהשעה 19:00.