项目作者: mkl04

项目描述 :
Reinforcement Learning's project
高级语言: Python
项目地址: git://github.com/mkl04/RL_project.git
创建时间: 2021-06-17T22:46:08Z
项目社区:https://github.com/mkl04/RL_project

开源协议:

下载


Reinforcement Learning’s project

Implementation of DQN and Double DQN for Atari Games.

Objective

Verify the advantage of the Double DQN algorithm over DQN (Deep Q Network) for two Atari games.

Environments

Only two games were chosen for training time reasons. These were Tennis and Space Invaders, since they demonstrated better results using Double DQN, as well as superiority with respect to human-level performance. The environments were taken from the open-source library gym, offered by OpenAI.

Solution

The code used is a customized version of the code shared on the Youtube channel brthor1.

Results

In the following link, it is shown a run for Space Invaders with a trained model for DQN.

DQN for Space Invaders game