Deep Reinforcement Learning in General Game Playing

Project description

Deep reinforcement learning (DRL) has made great achievements since proposed. The RL problem involves an agent interacting with an unknown or dynamic environment by executing a sequence of actions and learning from the observations and rewards received. In this project, we will examine how deep learning has been applied to the general decision-making problems in computer game playing, the challenges faced, and how they can be overcome. This project will implement the DRL models by starting with Deep-Q Learning (DQN) written in Python. More game environments from Gym or Mujoco could be chosen by interests.

Lecture notes: 01, 02, 03, 04, 05, 06.