Options of Interest: Temporal Abstraction with Interest Functions

TL;DR

This paper introduces interest functions for temporal abstraction in reinforcement learning, enabling better option discovery and interpretability through a gradient-based learning algorithm and new architecture.

Contribution

It generalizes initiation sets using interest functions, proposes a gradient-based learning algorithm, and develops an interest-option-critic architecture for improved temporal abstraction.

Findings

Interest functions improve option learning in discrete and continuous environments.

The approach yields interpretable and reusable temporal abstractions.

Quantitative and qualitative results demonstrate the method's effectiveness.

Abstract

Temporal abstraction refers to the ability of an agent to use behaviours of controllers which act for a limited, variable amount of time. The options framework describes such behaviours as consisting of a subset of states in which they can initiate, an internal policy and a stochastic termination condition. However, much of the subsequent work on option discovery has ignored the initiation set, because of difficulty in learning it from data. We provide a generalization of initiation sets suitable for general function approximation, by defining an interest function associated with an option. We derive a gradient-based learning algorithm for interest functions, leading to a new interest-option-critic architecture. We investigate how interest functions can be leveraged to learn interpretable and reusable temporal abstractions. We demonstrate the efficacy of the proposed approach through quantitative and qualitative results, in both discrete and continuous environments.