Module reinforcement_learner 

Reinforcement learning agents — Q-learning and SARSA for discrete action spaces.

Provides ReinforcementLearner with support for:

• Standard Q-learning (off-policy TD(0))
• SARSA (on-policy TD(0))
• Double Q-learning (reduces overestimation bias)
• Epsilon-greedy, greedy, and random exploration policies
• Per-step and per-episode statistics

All random number generation uses an inline xorshift64 PRNG seeded at construction time so the implementation is dependency-free and deterministic.

Structs

ActionId

Opaque identifier for an action.

Experience

A single (s, a, r, s’, done) transition.

ReinforcementLearner

Tabular reinforcement-learning agent supporting Q-learning, SARSA, and Double Q-learning.

RlStats

Aggregate statistics snapshot.

StateId

Opaque identifier for an environment state.

Enums

Policy

Action-selection policy.

RlAlgorithm

Reinforcement learning algorithm variant.

RlError

Errors that can be returned by reinforcement learning operations.