Approximate Value Function Methods

Instead of using tables (tabular methods) to represent value functions, for task with large state and action spaces approximate representations of value functions based on parametrizations should be used, especially if tabular methods render infeasible regarding memory consumption and computation times.

• Principal Approach
• Policy Evaluation
• Policy Improvement
• Deep Q-Learning

Related projects:

• Cart Pole

  • The Cart Pole Environment

  • SARSA

  • Deep Q-Learning