B0225
Title: Online data-driven decision-making in unknown continuous environments
Authors: Mohamad Kazem Shirani Faradonbeh - Southern Methodist University (United States) [presenting]
Abstract: One of the most popular dynamical models for continuous environments is that of linear stochastic differential equations. A widely applicable problem is learning to decide optimally to minimize a cost function when the true dynamics parameters are unknown. Implementable data-driven online learning algorithms are presented that learn the optimal decisions quickly via interacting with the environment. In fact, the proposed algorithm efficiently balances exploration versus exploitation by carefully randomizing the parameter estimates, such that its regret grows as the square root of time multiplied by the number of parameters. Theoretical performance analysis as well as experiments for learning to control an airplane will be presented to show efficiency.
