Udacity Deep Reinforcement Learning Nanodegree

This repository contains exercises and projects I've worked on for the nanodegree.

These are my reports for the three main projects, trained using Unity ML-Agents.

Navigation	Continuous Control	Collaboration and Competition
Implemented with a Dueling DDQN. See the README and the report.	Implemented with A2C-GAE. See the README and the report.	Implemented with MADDPG and self-play. See the README and the report.

Also, here's a list of exercise notebooks trained on a few OpenAI Gym environments.

• Monte Carlo methods
• Temporal difference methods (SARSA, Q-Learning, Expected-Sarsa)
• Discretization and tile coding (Q-Learning)
• Deep Q-Network
• Policy based methods (Hill Climbing, Cross Entropy Method)
• Policy gradient methods (REINFORCE, PPO)

Some python scripts derived from the notebooks are provided as well.