Timezone: »

 
Agent-Controller Representations: Principled Offline RL with Rich Exogenous Information
Riashat Islam · Manan Tomar · Alex Lamb · Hongyu Zang · Yonathan Efroni · Dipendra Misra · Aniket Didolkar · Xin Li · Harm Van Seijen · Remi Tachet des Combes · John Langford
Event URL: https://openreview.net/forum?id=0pFzg-8y-o »

Learning to control an agent from data collected offline in a rich pixel-based visual observation space is vital for real-world applications of reinforcement learning (RL). A major challenge in this setting is the presence of input information that is hard to model and irrelevant to controlling the agent. This problem has been approached by the theoretical RL community through the lens of exogenous information, i.e, any control-irrelevant information contained in observations. For example, a robot navigating in busy streets needs to ignore irrelevant information, such as other people walking in the background, textures of objects, or birds in the sky. In this paper, we focus on the setting with visually detailed exogenous information, and introduce new offline RL benchmarks offering the ability to study this problem. We find that contemporary representation learning techniques can fail on datasets where the noise is a complex and time dependent process, which is prevalent in practical applications. To address these, we propose to use multi-step inverse models, which have seen a great deal of interest in the RL theory community, to learn Agent-Controller Representations for Offline-RL (ACRO). Despite being simple and requiring no reward, we show theoretically and empirically that the representation created by this objective greatly outperforms baselines.

Author Information

Riashat Islam (MILA/McGill)
Manan Tomar (University of Alberta)
Alex Lamb (Universite de Montreal)
Hongyu Zang (Beijing Institute of Technology)
Yonathan Efroni (Microsoft Research, New York)
Dipendra Misra
Aniket Didolkar (University of Montreal)
Xin Li (Beijing Institute of Technology)
Harm Van Seijen (Microsoft Research)
Remi Tachet des Combes (Microsoft Research Montreal)
John Langford (Microsoft Research)

John Langford is a machine learning research scientist, a field which he says "is shifting from an academic discipline to an industrial tool". He is the author of the weblog hunch.net and the principal developer of Vowpal Wabbit. John works at Microsoft Research New York, of which he was one of the founding members, and was previously affiliated with Yahoo! Research, Toyota Technological Institute, and IBM's Watson Research Center. He studied Physics and Computer Science at the California Institute of Technology, earning a double bachelor's degree in 1997, and received his Ph.D. in Computer Science from Carnegie Mellon University in 2002. He was the program co-chair for the 2012 International Conference on Machine Learning.

More from the Same Authors