/tictactoe_agent

A reinforcement learning agent that learns and plays tic-tac-toe.

Primary LanguagePython

Tic Tac Toe Game

This repository trains an agent in the Tic Tac Toe game using Reinforcement Learning.

The first step is to train the agent to learn to play the game. After the training, the human can play against the agent.

To play the game, you need to have Docker installed.

  • Run the "run.sh" script
  • The board indices start from 0
  • The 1st square of the board is [0,0]

You can change the parameters of the agent by passing them as arguments to the script.

For example: "./run.sh 0.1, 0.5, 0.01, 10000"

Parameters

  1. Epsilon: The epsilon value. How much will the agent select a random action. Default: 0.1. Range: 0-1
  2. Alpha: The alpha value. The learning rate. Default: 0.5
  3. Decay: The decay value. The value by which the epsilon will decrease at each iteration. Default: 0.01
  4. Epochs: The number of iterations. Default: 10000