next up previous
Next: Policy Gradient Methods: Swimmer Up: MLB_Exercises_2010 Previous: RL application II: Function

RL application III: Self-play [3* P]

Learn a successful game strategy for a simple two-player game via self-play. The choice of the game and the learning algorithm is up to you. Suggestions for two-player games are Tic-Tac-Toe, Blackjack or Nim. If you want to choose you own game please send an email to before you start with the implementation. Implement the RL learning algorithm and the game environment in MATLAB using the Reinforcement Learning MATLAB Toolbox 6. Evaluate the performance of your method. Document the implementation and the results in such a way that anybody can reproduce them effortless.

Haeusler Stefan 2011-01-25