Empower - The Campaign for Northeastern University


ECE PhD Defense: "Function Approximation-based Reinforcement Learning for Large-Scale Problem Domains," Wei Li

Abstract: A central challenge in reinforcement learning (RL) is to maintain the RL agent’s performance in very large, continuous state spaces. Function approximation is the primary tool to solve the performance degradation issue when implementing RL algorithms in large-scale, continuous problems. In such problem domains, the size of the state space can grow exponentially with the number of state variables. As a result, the size of the table needed to store the state or state-action values and the resulting training time to explore the state-action space can limit the complexity of the...

Wednesday, November 28, 2018 at 2:00pm





Google Calendar iCal Outlook

Recent Activity