Browsing: Future reward estimation

Researchers have introduced Temporal Reward Decomposition (TRD) to enhance explainability in reinforcement learning by predicting the next N expected rewards, revealing when and what…