Learning Curriculum Policies for Reinforcement Learning

TL;DR

This paper introduces a method to automatically learn curriculum policies for reinforcement learning by modeling the task sequencing as a Markov Decision Process, enabling faster training of agents on complex tasks.

Contribution

It extends existing curriculum design models to handle multiple transfer algorithms and demonstrates learning curriculum policies from experience.

Findings

Curriculum policies trained with our method accelerate agent learning.

Our approach outperforms existing curriculum methods in speed.

The method is effective across multiple domains and agents.

Abstract

Curriculum learning in reinforcement learning is a training methodology that seeks to speed up learning of a difficult target task, by first training on a series of simpler tasks and transferring the knowledge acquired to the target task. Automatically choosing a sequence of such tasks (i.e. a curriculum) is an open problem that has been the subject of much recent work in this area. In this paper, we build upon a recent method for curriculum design, which formulates the curriculum sequencing problem as a Markov Decision Process. We extend this model to handle multiple transfer learning algorithms, and show for the first time that a curriculum policy over this MDP can be learned from experience. We explore various representations that make this possible, and evaluate our approach by learning curriculum policies for multiple agents in two different domains. The results show that our method produces curricula that can train agents to perform on a target task as fast or faster than existing methods.