SUNMONTUEWEDTHUFRISAT
Events for the 4th week of March
-
ECE Seminar: Safe Deep Learning in the Feedback Loop: A Robust Control Approach
Mon, Mar 23, 2020 @ 11:00 AM - 12:00 PM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Mahyar Fazlyab, Postdoctoral Researcher, Dept of ESE, University of Pennsylvania
Talk Title: Safe Deep Learning in the Feedback Loop: A Robust Control Approach
Abstract: Despite high-profile advances in various decision-making and classification tasks, Deep Neural Networks (DNNs) face several fundamental challenges that limit their adoption in physical or safety-critical domains. In particular, DNNs can be vulnerable to adversarial attacks and input perturbations. This issue becomes even more pressing when DNNs are used in closed-loop systems, where a small perturbation (caused by, for example, noisy measurements, uncertain initial conditions, or disturbances) can substantially impact the system being controlled. Therefore, it is of utmost importance to develop tools that can provide useful certificates of stability, safety, and robustness for DNN-driven systems.
In this talk, I will present a new framework, rooted in convex optimization and robust control, for safety verification and robustness analysis of DNNs based on semidefinite programming. The main idea is to abstract the original, nonlinear, hard-to-analyze neural network by a Quadratically-Constrained Linear Network (QCLN), in which the nonlinear components (e.g., the activation functions) are described by the quadratic constraints that all their input-output instances satisfy. This abstraction allows us to analyze various properties of DNNs (safety, local and global robustness, etc.) using semidefinite programming.
Biography: Mahyar Fazlyab received the Bachelor's and Master's degrees in mechanical engineering from Sharif University of Technology, Tehran, Iran, in 2010 and 2013, respectively. He earned a Master's degree in statistics and a Ph.D. degree in Electrical and Systems Engineering (ESE) from the University of Pennsylvania (UPenn), Philadelphia, PA, USA, in 2018. Currently, he is a Postdoctoral Researcher at UPenn. His research interests are at the intersection of optimization, control, and machine learning. His current work focuses on developing optimization-based methods for safety verification of learning-enabled control systems. Dr. Fazlyab won the Joseph and Rosaline Wolf Best Doctoral Dissertation Award in 2019, awarded by the ESE Department at UPenn.
Host: Mihailo Jovanovic, mihailo@usc.edu, 213.740.4474
Webcast: https://usc.zoom.us/j/871407253WebCast Link: https://usc.zoom.us/j/871407253
Audiences: Everyone Is Invited
Contact: Mayumi Thrasher
-
ECE Seminar: Reliability, Equity, and Reproducibility in Modern Machine Learning
Tue, Mar 24, 2020 @ 11:00 AM - 12:00 PM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Yaniv Romano, Postdoctoral Scholar, Dept of Statistics, Stanford University
Talk Title: Reliability, Equity, and Reproducibility in Modern Machine Learning
Abstract: Modern machine learning algorithms have achieved remarkable performance in a myriad of applications, and are increasingly used to make impactful decisions in the hiring process, criminal sentencing, healthcare diagnostics and even to make new scientific discoveries. The use of data-driven algorithms in high-stakes applications is exciting yet alarming: these methods are extremely complex, often brittle, notoriously hard to analyze and interpret. Naturally, concerns have raised about the reliability, fairness, and reproducibility of the output of such algorithms. This talk introduces statistical tools that can be wrapped around any "black-box" algorithm to provide valid inferential results while taking advantage of their impressive performance. We present novel developments in conformal prediction and quantile regression, which rigorously guarantee the reliability of complex predictive models, and show how these methodologies can be used to treat individuals equitably. Next, we focus on reproducibility and introduce an operational selective inference tool that builds upon the knockoff framework and leverages recent progress in deep generative models. This methodology allows for reliable identification of a subset of important features that is likely to explain a phenomenon under-study in a challenging setting where the data distribution is unknown, e.g., mutations that are truly linked to changes in drug resistance.
Biography: Yaniv Romano is a postdoctoral scholar in the Department of Statistics at Stanford University, advised by Prof. Emmanuel Candes. He earned his Ph.D. and M.Sc. degrees in 2017 from the Department of Electrical Engineering at the Technion-”Israel Institute of Technology, under the supervision of Prof. Michael Elad. Before that, in 2012, Yaniv received his B.Sc. from the same department. His research spans the theory and practice of selective inference, sparse approximation, machine learning, data science, and signal and image processing. His goal is to advance the theory and practice of modern machine learning, as well as to develop statistical tools that can be wrapped around any data-driven algorithm to provide valid inferential results. Yaniv is also interested in image recovery problems: the super-resolution technology he invented together with Dr. Peyman Milanfar is being used in Google's flagship products, increasing the quality of billions of images and bringing significant bandwidth savings. In 2017, he constructed with Prof. Michael Elad a MOOC on the theory and practice of sparse representations, under the edX platform. Yaniv is a recipient of the 2015 Zeff Fellowship, the 2017 Andrew and Erna Finci Viterbi Fellowship, the 2017 Irwin and Joan Jacobs Fellowship, the 2018-2020 Zuckerman Postdoctoral Fellowship, the 2018-2020 ISEF Postdoctoral Fellowship, the 2018-2020 Viterbi Fellowship for nurturing future faculty members, Technion, and the 2019-2020 Koret Postdoctoral Scholarship, Stanford University. Yaniv was awarded the 2020 SIAG/IS Early Career Prize.
Host: Salman Avestimehr, avestime@usc.edu
Webcast: https://usc.zoom.us/j/782728120WebCast Link: https://usc.zoom.us/j/782728120
Audiences: Everyone Is Invited
Contact: Mayumi Thrasher