Reinforcement Learning

• If you played a game with a computer agent, could it learn to improve its performance.
• If it took feedback from the environment, and used that to choose between its rules, it would be using reinforcement learning.
• Roman and I did this with CANT to learn to choose the correct rule.
• Bucket brigade and delay of reward
• It can also be used in game theory. It fits in with utility theory quite nicely.