Sign-separated finite-time bounds for constant-step-size Q-learning certify a slower exponential envelope for positive errors than for negative errors, reflecting max-induced overestimation.
Finite-time analysis of asynchronous Q-learning under diminishing step-size from control-theoretic view.IEEE Access, 12:149916–149939
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it