Posted November 8, 2019
Improving Generalization and Abstraction in Deep Reinforcement Learning
Team Members and Titles

RPI—PI: Chris R. SimsAssistant Professor, Cognitive Science Department, Rensselaer Polytechnic Institute, Troy, NY

IBM—PI: Tim KlingerResearch Staff Member, IBM Research AI Learning, Thomas J. Watson Research Center, Yorktown Heights, NY

 

Additional key collaborators:

Gerald TesauroPrincipal Research Staff Member, Thomas J. Watson Research Center, Yorktown Heights, NY

Matt RiemerResearch Engineer, Thomas J. Watson Research Center, Yorktown Heights, NY

Miao LiuResearch Staff Member, Thomas J. Watson Research Center, Yorktown Heights, NY

Summary

This project seeks to develop novel methods for improving the generalization of learning within the context of computational reinforcement learning, by leveraging theoretical results from cognitive science research.

Key Findings

* Developed novel machine learning algorithms for reinforcement learning and hierarchical reinforcement learning

* Experimentally evaluated generalization and abstraction in both small scale and large scale learning environments

Project Start Date
Project End Date