Meta reinforcement learning as task inference

TL;DR

This paper introduces a method for meta reinforcement learning that infers task structure through belief estimation, enabling efficient learning in complex environments with limited prior knowledge.

Contribution

It proposes a novel approach to separately learn policy and task belief using privileged information, improving performance in standard and complex meta-RL tasks.

Findings

Effective in standard meta-RL environments

Performs well in complex continuous control tasks

Utilizes privileged information for belief estimation

Abstract

Humans achieve efficient learning by relying on prior knowledge about the structure of naturally occurring tasks. There is considerable interest in designing reinforcement learning (RL) algorithms with similar properties. This includes proposals to learn the learning algorithm itself, an idea also known as meta learning. One formal interpretation of this idea is as a partially observable multi-task RL problem in which task information is hidden from the agent. Such unknown task problems can be reduced to Markov decision processes (MDPs) by augmenting an agent's observations with an estimate of the belief about the task based on past experience. However estimating the belief state is intractable in most partially-observed MDPs. We propose a method that separately learns the policy and the task belief by taking advantage of various kinds of privileged information. Our approach can be very effective at solving standard meta-RL environments, as well as a complex continuous control environment with sparse rewards and requiring long-term memory.