Speaker: Linfeng Zhao Northeastern University
Time: 2024-01-04 16:00-2024-01-04 17:00
Venue: C19-2 or Tencent Meeting ID: 398-077-2882, code: 095015

Abstract:

This talk will cover the integration of geometric structures in robotic learning and planning. Focused on the challenges of robotic applications and their algorithms, it explores how geometric structures act as inductive biases to enhance learning and planning. The discussion will cover Linfeng’s recent research contributions, including methods for learning representations and integrating symmetry and compositionality, demonstrating their impact on the efficiency and generalizability of robotic systems. This presentation aims to shed light on the potential of geometric structures in more generalizable and scalable intelligent robots.

Short Bio:

Linfeng Zhao is a Ph.D. candidate at Northeastern University, advised by Prof. Lawson Wong and working with Prof. Robin Walters, focusing on machine learning, robotics, and artificial intelligence. His research centers on enhancing robotic decision-making, particularly for long-horizon mobile-manipulation robots. Linfeng has recently worked with Jenny Barry at Boston Dynamics AI Institute on apply his research on Boston Dynamics Spot. Before, Linfeng worked at Amazon Science, Microsoft Research Asia, and Hao Su’s lab at UC San Diego. He has published at venues like ICML, NeurIPS, and ICLR.