MachineLearning/sarsa library

SARSA (on-policy TD control)

Simple SARSA implementation for discrete state and action spaces. This algorithm follows the on-policy update rule and exposes an epsilon-greedy action selection. The API mirrors the tabular Q-Learning class so users can swap algorithms easily in examples and tests.

Contract:

• states/actions: integers
• update(state, action, reward, nextState, nextAction)

Classes

SARSA

SARSA (on-policy TD control) with optional schedules and helpers