This is an implementation of the AlphaZero algorithm with TensorFlow.js. You can use it to play the simple board game gobang.

😼 😺 😿 😹 😾

npm run start

npm run train

This model has been trained for about 3000 innings in total. The following is the match situation of self play in the 2800 innings. It can be seen that the AI program has learned how to attack and defend, and it is believed that it will become more comprehensive and flexible after more rounds of training.

Tips: When playing chess, the AI speed is relatively slow, please be patient for a moment.