\textit{SQT} -- \textit{std} $Q$-target

\textit{Std} $Q$-target is a \textit{conservative}, actor-critic, ensemble, $Q$-learning-based algorithm, which is based on a single key $Q$-formula: $Q$-networks standard deviation, which is an "uncertainty penalty", and, serves as a minimalistic solution to the problem of \textit{overestimation} bias. We implement \textit{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 \textit{SQT}'s $Q$-target formula superiority over \textit{TD3}'s $Q$-target formula as a \textit{conservative} solution …

actor actor-critic algorithm algorithms art bias code cs.ai cs.lg ddpg deviation ensemble key networks solution sota standard state test uncertainty