Regularized Emphatic Temporal-Difference Learning: Stability under Constant Stepsizes
This work introduces regularized emphatic TD (RETD), a normalized first-order post-shock repair that leaves the trace and importance ratios unchanged, stores the emphatic TD signal in a leaky scalar state, and releases a delayed correction.
Xing-Guo Chen, Zhao-Hui Wu, Ji Ye et al.
· 0 citations