Reinforcement learning

From Wikiversity
Jump to: navigation, search

Reinforcement learning

What is Reinforcement Learning[edit]

"Reinforcement learning (RL) is learning from interaction with an environment, from the consequences of action, rather than from explicit teaching." -- Rich Sutton

Evaluative Feedback (Chapter 2)[edit]

Softmax Action Selection[edit]

Softmax action selection is the way to maintain exploration and exploitation balance. The softmax policy will choose action a on period t with probability:

Reinforcement Learning Problems (Sutton and Barto Chapter 3)[edit]

Dynamic Programming (Sutton and Barto Chapter 4)[edit]

Monte Carlo Methods (Sutton and Barto Chaper 5)[edit]

Temporal-Difference Learning (Sutton and Barto Chapter 6)[edit]

Eligibility Traces (Sutton and Barto Chapter 7)[edit]

Related Terms[edit]

  • Machine Learning
  • Adaptive Dynamic Programming
  • Markov Decision Process

References[edit]