English

SQT -- std $Q$-target

Machine Learning 2024-06-04 v3 Artificial Intelligence

Abstract

Std QQ-target is a conservative, actor-critic, ensemble, QQ-learning-based algorithm, which is based on a single key QQ-formula: QQ-networks standard deviation, which is an "uncertainty penalty", and, serves as a minimalistic solution to the problem of overestimation bias. We implement SQT on top of TD3/TD7 code and test it against the state-of-the-art (SOTA) actor-critic algorithms, DDPG, TD3 and TD7 on seven popular MuJoCo and Bullet tasks. Our results demonstrate SQT's QQ-target formula superiority over TD3's QQ-target formula as a conservative solution to overestimation bias in RL, while SQT shows a clear performance advantage on a wide margin over DDPG, TD3, and TD7 on all tasks.

Cite

@article{arxiv.2402.05950,
  title  = {SQT -- std $Q$-target},
  author = {Nitsan Soffair and Dotan Di-Castro and Orly Avner and Shie Mannor},
  journal= {arXiv preprint arXiv:2402.05950},
  year   = {2024}
}
R2 v1 2026-06-28T14:43:21.108Z