News & EventsDepartment Events & Announcements
Events
-
Jun1
EVENT DETAILS
lessLivestream:
https://northwestern.hosted.panopto.com/Panopto/Pages/Viewer.aspx?id=7a07cf41-b0cc-4b80-8291-abc9015d4ccbTitle:
Computational and Statistical Efficiency of Policy OptimizationAbstract
Coupled with powerful function approximators such as neural networks, policy optimization plays a key role in the tremendous empirical successes of deep reinforcement learning. In sharp contrast, the theoretical understandings of policy optimization remain rather limited from both the computational and statistical perspectives. From the perspective of computational efficiency, it remains unclear whether policy optimization converges to the globally optimal policy in a finite number of iterations, even given infinite data. From the perspective of statistical efficiency, it remains unclear how to attain the globally optimal policy with a finite regret or sample complexity.To address the computational question, I will show that, under suitable conditions, natural policy gradient/proximal policy optimization/trust-region policy optimization (NPG/PPO/TRPO) converges to the globally optimal policy at a sublinear rate, even when it is coupled with neural networks. To address the statistical question, I will present an optimistic variant of NPG/PPO/TRPO, namely OPPO, which incorporates exploration in a principled manner and attains a \sqrt{T}-regret.
(Joint work with Qi Cai, Chi Jin, Jason Lee, Boyi Liu, Zhuoran Yang)
Biography
Zhaoran Wang is an assistant professor at Northwestern IEMS, working at the interface of machine learning, statistics, and optimization. He is the recipient of the AISTATS (Artificial Intelligence and Statistics Conference) notable paper award, ASA (American Statistical Association) best student paper in statistical learning and data mining, INFORMS (Institute for Operations Research and the Management Sciences) best student paper finalist in data mining, and the Microsoft fellowship.TIME Monday, June 1, 2020 at 12:00 PM - 1:00 PM
LOCATION Zoom, Mudd Hall ( formerly Seeley G. Mudd Library) map it
CONTACT Pam Villalovoz pmv@northwestern.edu EMAIL
CALENDAR Department of Computer Science
-
Jun3
EVENT DETAILS
lessTitle:
Computational Law - The Cop in the BackseatAbstract
Computational Law is a branch of Legal Informatics concerned with the codification of rules and regulations in computable form. From a philosophical point of view, Computational Law is most closely associated with the formalist school of jurisprudence. From a practical point of view, it is the basis for the implementation and deployment of computer systems capable of doing useful legal calculations, such as compliance checking, legal planning, and so forth. Computational Law has the potential to dramatically change the legal profession, improving the quality and efficiency of legal services and possibly disrupting the way law firms do business. More broadly, the technology has the potential to bring legal understanding and legal tools to everyone in society, not just legal professionals, thus enhancing access to justice and improving the legal system as a whole.Biography
Michael Genesereth is a professor in the Computer Science Department at Stanford University and a professor by courtesy in the Stanford Law School. He received his Sc.B. in Physics from M.I.T. and his Ph.D. in Applied Mathematics from Harvard University. Genesereth is most known for his work on Computational Logic and applications of that work in Enterprise Management, Computational Law, and General Game Playing. He is the director of the Logic Group at Stanford and the founder and research director of CodeX - the Stanford Center for Legal Informatics.TIME Wednesday, June 3, 2020 at 12:00 PM - 1:00 PM
LOCATION Zoom, Mudd Hall ( formerly Seeley G. Mudd Library) map it
CONTACT Pam Villalovoz pmv@northwestern.edu EMAIL
CALENDAR Department of Computer Science