Nov 30, 2023

Huazhe Xu

Huazhe Xu is an Assistant Professor at the Institute for Interdisciplinary Information Sciences, Tsinghua University. He completed his postdoctoral studies at Stanford University and obtained his Ph.D. from the University of California, Berkeley. His research focuses on the theory, algorithms, and applications of Embodied AI, including deep reinforcement learning, robotics, and sensorimotor control. He has made significant contributions to addressing core challenges in Embodied AI, such as low data efficiency and weak generalization capabilities. Dr. Xu has published over 40 papers in top-tier academic conferences, and his notable works have been featured in media outlets such as MIT Tech Review and Stanford HAI.

Abstract

Reinforcement learning (RL) has emerged as a powerful paradigm for training intelligent agents through trial and error. This talk provides an overview of RL algorithms, covering the basics and highlighting the challenges associated with their application. We delve into the cutting-edge methods that address the crucial aspects of sample efficiency and generalization ability in RL. These advancements enable agents to learn efficiently from limited data and generalize their knowledge to new situations. Additionally, we explore the practical aspects of real-world deployment, considering the complexities and considerations involved in integrating RL agents into real-world applications. Join us for an engaging discussion on the latest developments in RL, along with insights into optimizing efficiency, enhancing generalization, and overcoming challenges for successful real-world deployment.