:::

[2019-11-13] Dr. Yi Ouyang, Preferred Networks (PFN), "Improving learning efficiency in dynamical systems with model-based approaches"

專題討論演講公告
Poster:Post date:2019-10-23
Title: Improving learning efficiency in dynamical systems with model-based approaches
Date: 2019-11-13 2:20pm-3:30pm
Location: R210, CSIE
Speaker: Dr. Yi Ouyang, Preferred Networks (PFN)
Hosted by: Prof. Yen-Huan Li
 
 

Abstract:

 
Recent advances in data-driven approaches have shown impressive results in prediction and control of dynamical systems. In particular, deep neural networks provide accurate predictions and superhuman performance in some application domains. However, pure data-driven methods generally require a huge amount of data, and the learned models often have limited generalizability. These issues restrict the applicability of pure data-driven methods in many real world dynamical systems where data collection is expensive and generalization is critical. Model-based approaches aim to solve the sample-efficiency and generalization challenges by constructing models to utilize prior knowledge. In this talk, I will share the idea of combining data-driven and analytical models to improve generalization and online adaptation in predicting physical interactions. I will also discuss model-based reinforcement learning methods and new approaches to improve sample efficiency and performance in dynamical environments.
 
 
Biography:
 
Yi Ouyang is a researcher at Preferred Networks (PFN). His research focuses on reinforcement learning, multi-agent systems, robot manipulation and navigation, decentralized control, and stochastic games. Prior to joining PFN, he was a postdoctoral scholar at the University of California, Berkeley, California from 2017 to 2018, and a postdoctoral scholar at the University of Southern California, Los Angeles, California from 2016 to 2017. He received the B.S. degree in Electrical Engineering from the National Taiwan University, Taipei, Taiwan in 2009, and the Ph.D. degree in Electrical Engineering and Computer Science at the University of Michigan in 2015.



 
Last modification time:2019-10-23 AM 9:04

cron web_use_log