Smart Sampling: Self-Attention and Bootstrapping for Improved Ensembled Q-Learning

TL;DR

This paper introduces a novel ensemble Q-learning method called Smart Sampling that uses self-attention and bootstrapping to improve sample efficiency, reduce bias, and enhance performance, especially in low UTD scenarios.

Contribution

The paper proposes a simple yet effective method integrating self-attention and bootstrapping into ensemble Q-learning, outperforming existing methods like REDQ and DroQ.

Findings

Improves Q prediction accuracy and sample efficiency.

Reduces normalized bias and its standard deviation.

Performs well with low update-to-data ratios.

Abstract

We present a novel method aimed at enhancing the sample efficiency of ensemble Q learning. Our proposed approach integrates multi-head self-attention into the ensembled Q networks while bootstrapping the state-action pairs ingested by the ensemble. This not only results in performance improvements over the original REDQ (Chen et al. 2021) and its variant DroQ (Hi-raoka et al. 2022), thereby enhancing Q predictions, but also effectively reduces both the average normalized bias and standard deviation of normalized bias within Q-function ensembles. Importantly, our method also performs well even in scenarios with a low update-to-data (UTD) ratio. Notably, the implementation of our proposed method is straightforward, requiring minimal modifications to the base model.