Talk Abstract: The talk will discuss the challenges of certifying the safety and robustness properties of machines that learn. I will describe our work that uses convex relaxations and set partitioning to simplify the analysis of highly nonlinear neural networks used across AI. These analysis tools led to the first framework for deep reinforcement learning that is certifiably robust to adversarial attacks and noisy sensor data. The tools also enable reachability analysis -- the calculation of all states that a system could reach in the future -- for systems that employ neural networks in the feedback loop, which provides another notion of safety for learning machines that interact with uncertain environments.

Michael Everett is currently a Visiting Faculty Researcher with Google’s People + AI Research (PAIR) team, he will be joining Northeastern University in January 2023 as an Assistant Professor, with a joint appointment in the Department of Electrical & Computer Engineering and the Khoury College of Computer Sciences. His research lies at the intersection of robotics, deep learning, and control theory, with the goal of developing certifiable learning machines. Previously, He was a Research Scientist and Postdoctoral Associate at the MIT Department of Aeronautics and Astronautics, working on the DARPA RACER program, the ARL SARA program, and advancing the field of certifiable learning. He received the PhD (2020), SM (2017), and SB (2015) degrees from MIT in Mechanical Engineering.

свернуть