Reinforcement Learning

A framework for setting up and training agents using reinforcement learning.

Key concepts:

Environment: A representation of some problem or scenario.

State: Information needed to define the current state of the Environment.

Agent: An object that is capable of deciding what Action to take in order to modify the State according to some strategy.

Reward: The thing the Agent tries to maximize through its actions.

Action: A change in State that an Agent can effect.

Scorer: The thing that determines what the Reward is for a given State.

Updater: The thing that decides how Actions modify State.