Policy Search with Rare Significant Events: Choosing the Right Partner to Cooperate with

TL;DR

This paper investigates reinforcement learning in scenarios with rare significant events, demonstrating that evolutionary strategies outperform gradient methods in such settings, especially with continuous states and actions.

Contribution

It compares gradient and direct policy search methods, highlighting the robustness of evolutionary algorithms when significant events are scarce.

Findings

Gradient methods struggle with rare significant events.

Evolutionary strategies are invariant to event rarity.

Direct policy search methods outperform gradient methods in this context.

Abstract

This paper focuses on a class of reinforcement learning problems where significant events are rare and limited to a single positive reward per episode. A typical example is that of an agent who has to choose a partner to cooperate with, while a large number of partners are simply not interested in cooperating, regardless of what the agent has to offer. We address this problem in a continuous state and action space with two different kinds of search methods: a gradient policy search method and a direct policy search method using an evolution strategy. We show that when significant events are rare, gradient information is also scarce, making it difficult for policy gradient search methods to find an optimal policy, with or without a deep neural architecture. On the other hand, we show that direct policy search methods are invariant to the rarity of significant events, which is yet another confirmation of the unique role evolutionary algorithms has to play as a reinforcement learning method.