MachineLearning/q_learning library

Q-Learning (tabular)

A compact, well-documented tabular Q-Learning implementation suitable for discrete state/action spaces. This class intentionally keeps the API simple (states and actions are integers) so it can be composed into larger agents or used in unit tests and examples.

Public contract:

  • Inputs: state (int), action (int), reward (double), nextState (int)
  • Storage: internal Q-table of shape (nStates x nActions)
  • Outputs: Q-values, action selection via epsilon-greedy

Classes

QLearning
Tabular Q-Learning (advanced)