/Deep-Reinforcement-Learning-Course-Udacity

This repository contains my solutions to the projects in the Deep Reinforcement Learning Nanodegree from Udacity.

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

Deep Reinforcement Learning Nanodegree at Udacity

This repository presents my solutions to the projects for the Udacity Nanodegree.

Projects

Project 1: Navigation

In this project an agent is trained to collect yellow bananas (+1) while avoiding blue bananas (-1) in a 3D world.

The solution employs a Deep Q-Network (DQN) and learns to collect an average score of more than 13 after 500 episodes, but performance improves up to around 900 episodes.

The specific details and solution to the project can be found in the subfolder "Project 1 - Navigation with DQN Agent".

Getting Started

Specific details for running each project code can be found under each project folder.

In general the following frameworks are used:

  • Python 3.6
  • PyTorch 0.4.0
  • Unity ML-Agents 0.4.0

Other project-specific dependencies are noted in the README for each project.

License

GPL-3.0

Author

Simon Bøgh