Nicolai Czempin 73495ae7e4 cleanup
2017-04-04 23:07:45 +02:00
2017-04-04 23:07:45 +02:00
2017-03-31 00:36:55 +02:00
2017-03-31 00:36:55 +02:00
2017-04-03 21:55:40 +02:00
2017-04-03 23:51:35 +02:00

gym-tic-tac-toe

An example of a custom environment for https://github.com/openai/gym.

I want to try out self-play in a Reinforcement Learning context. Rather than the board game environments on openai/gym right now, which are "single-player" by providing a built-in opponent, I want to create an agent that learns a strategy by playing against itself, so it will try to maximize the reward for "player 1" and minimize it for "player 2".

The canonical example of a simple two player game is Tic Tac Toe, also known as Noughts and Crosses.

Description
A simple two-player environment for openai/gym
Readme 49 KiB
Languages
Python 100%