RL$^3$: Boosting Meta Reinforcement Learning via RL inside RL$^2$

Abhinav Bhatia, Samer B. Nashed, Shlomo Zilberstein

Keywords: meta-reinforcement learning

TL;DR: We propose incorporating action-values, learned online via traditional RL, as inputs to meta-RL and show that this improves earned cumulative reward over longer adaptation periods and generalizes better to out-of-distribution tasks.

Abstract: Meta reinforcement learning (meta-RL) methods such as RL$^2$ have emerged as promising approaches for learning data-efficient RL algorithms tailored to a given task distribution. However, they show poor asymptotic performance and struggle with out-of-distribution tasks because they rely on sequence models, such as recurrent neural networks or transformers, to process experiences rather than summarize them using general-purpose RL components such as value functions. In contrast, traditional RL algorithms are data-inefficient as they do not use domain knowledge, but do converge to an optimal policy in the limit. We propose RL$^3$, a principled hybrid approach that incorporates action-values, learned per task via traditional RL, in the inputs to meta-RL. We show that RL$^3$ earns greater cumulative reward in the long term compared to RL$^2$ while drastically reducing meta-training time and generalizing better to out-of-distribution tasks. Experiments are conducted on both custom and benchmark discrete domains from the meta-RL literature that exhibit a range of short-term, long-term, and complex dependencies.

Submission Number: 52