Temporal-difference error mathematical notation

Temporal-difference error is a recurring research-paper notation family. Compares a one-step reward-plus-next-value target with the current value prediction to produce a signed learning signal.

Temporal-difference error: Compares a one-step reward-plus-next-value target with the current value prediction to produce a signed learning signal. Example: Measure how surprising the observed transition was to the current value estimate.