Introduction

This is the Kaggle competition regarding Game AI and Reinforcement Learning.

The project statement was to design an agent that would algorithmically play the most optimal move in the 2-player game Connect 4.

I made two agents for this - one using the techniques and algorithms learnt in the course AI: Search Methods for Problem Solving and another using Deep Reinforcement Learning.

The agent made with the first approach is documented in this section, the other one can be found here.

First I have used the minimax algorithm which looks at every possibility 5 moves ahead, then I have used Alpha-Beta pruning to optimize it by removing certain nodes from consideration without affecting the final evaluation, which means less possibilities are seen and the algorithm is faster.

The notebook I made for this is: notebook