Module utility 

Utility calculation algorithms for episode ranking

Ported from MemRL with adaptations for Smelt’s needs.

Structs

DecayParams

Parameters for exponential decay

PropagationResult

Result of a propagation run

UtilityRanker

Utility-based ranker for episodes

Functions

apply_decay

Apply exponential decay to a utility value

bellman_propagate

Run Bellman propagation to spread utility through the memory

temporal_credit_assignment

Run temporal credit assignment

wilson_score

Calculate Wilson score lower bound