UmbrellaRL-ctorrington 0.0.0

Reiforcement Learning package description.

Gradient-MC-1000-State-Random-Walk

Structure

• Agent

The Agent represents the logic for reinforcement learning. It contains functions for policy evaluation & calculating a State's value.

• Policy

The Policy object determines the actions taken by the agent. The Policy object must inherit from the Base Policy class.

• Environment

The Environment represents the State that the Agent is in. It describes which states are accessible following actions & what their probabilities are.

The Environment depends on the State Space.

• State Space

The State Space describes the states that the agent can be in. Each State is of type State.

The State Space depends on the State.

• State

The State represents a singular State that the agent can be in. The State contains the information that describes each State.