Smart Sampling: Self-Attention and Bootstrapping for Improved Ensembled Q-Learning
This work addresses sample efficiency for reinforcement learning practitioners, but it is incremental as it builds on existing ensemble Q-learning methods like REDQ and DroQ.
The paper tackles the problem of sample efficiency in ensemble Q-learning by integrating multi-head self-attention and bootstrapping, resulting in performance improvements over REDQ and DroQ, reduced bias and standard deviation in Q-function ensembles, and effective operation at low update-to-data ratios.
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.