Shihan Wang

Date: 16:00 – 16:30, Thursday, 14.10.2021
Location: MS Teams ICS Colloquium & Minnaert 2.02

Title:Learning from Interaction: The Challenge in Reinforcement Learning
Abstract:Reinforcement learning (RL) is concerned with how to sequentially interact with an environment to maximize a long-term goal. In contrast to other techniques in machine learning, RL focuses on learning from interactions, enabling intelligent systems to learn an optimal solution adaptively. In recent years, RL techniques have attracted attention and are used in a variety of real-world scenarios, such as robot navigation, mobile health intervention and autonomous driving. However, this interactive nature also poses various technical challenges for RL-based systems. In this talk, I will focus on two of those challenges, namely, the constraint of interaction frequency and partially observable environments. Several our on-going projects that address these challenges will be also discussed.