I am looking to create a challenging AI project.
There is an online card game that allows you to build a deck to play with out of randomly presented cards. You have to choose one each time out of a couple of cards. I want to build an AI that picks the cards as I would pick them myself using my knowledge of the game.
My guess is that I would have to create a system that can be trained based on “experience” (taught by training) and adapts its choices based on which previous cards it has already selected.
I am looking for the specific name for this type of AI so I can start finding some resources for it or similar projects if you know any.
1
I believe the General term that you are looking for is Reinforcement Learning: http://en.wikipedia.org/wiki/Reinforcement_learning. A popular model free form of this is called Q Learning.
A recent example usage of this was to beat that incredibly difficult FlappyBird game: http://sarvagyavaish.github.io/FlappyBirdRL/
1