Title:Revised rogressive-Hedging-Algorithm Based Two-layer Solution Scheme for Bayesian Reinforcement Learning
Presenter: Prof. Li Duan
Hosted by: Prof.Li Ping
Time:2019.9.26 16:00-18:00
Location:A1028
Abstract:Stochasticcontrol with both inherent random system noise and lack of knowledge on system parameters constitutes a fundamental challenge in reinforcement learning,especially under non-episodic setting. We propose a novel two-layer solution scheme to separate reducible system uncertainty from irreducible one at two layers (adopting time-composition based DP at the lower layer and the scenario-decomposition based progressive hedging algorithm at the upper layer) and to approximate the optimal policy directly. Applications in dynamic portfolio selection will be discussed.
