Temporal Difference Learning (value bootstrapping)
Definition
A method for handling long-horizon and sparse rewards by maintaining a value function predicting the long-term outcome; when intermediate progress changes that prediction, the change itself reinforces the action that caused it, allowing learning from intermediate steps without waiting for the final result.
Cast within
No topic-region cast yet — this appears once Temporal Difference Learning (value bootstrapping)'s compiled claims are aligned into a topic region's argument tree.
Claims mentioning Temporal Difference Learning (value bootstrapping) (3)
Your notes
Loading notes...