Stolle M, Precup D. Learning options in reinforcement learning. InInternational Symposium on abstraction, reformulation, and approximation 2002 Aug 2 (pp. 212-223). Springer, Berlin, Heidelberg.

No code.

Abstract

Why is it good?

Previous methods showed how to incorporate prior knowledge in AI, but how good options might be is difficult to identify.

Core Technology

Verification Method

Build Grid World with narrow gates so that robots learn to learn paths that choose those narrow gates.

Any Argument or Idea?

Next Paper to read

Details