Personal Work Files for Udacity's Deep Reinforcement Learning Nanodegree Program

Exercices

• • [x] 01. Dynamic Programming
• • [x] 02. Monte Carlo Control
• • [x] 03. Temporal Difference Control
• • [x] 04. Discretization
• • [x] 05. Tile Coding
• • [x] 06. Deep Q-Networks
• • [ ] 07. Robotics
• • [x] 08. Hill Climbing
• • [x] 09. Cross Entropy
• • [x] 10. REINFORCE
• • [ ] 11. Proximal Policy Optimization (PPO)
• • [x] 12. Deep Deterministic Policy Gradients (DDPG)
• • [ ] 13. Multi-Agent DDPG
• • [ ] 14. AlphaZero

Labs and Graded Projects

• • [ ] P0. Taxi - general RL.
• • [x] P1. Navigation - value-based methods.
• • [x] P2. Continuous Control - policy-based methods.
• • [x] P3. Collaboration and Competition - multi-agent RL.

Setup

TODO: Dockerfiles for CPU/GPU machines.

# Environment
conda create --name drlnd python=3.6
source activate drlnd

# OpenAI gym
git clone https://github.com/openai/gym.git
cd gym
pip install -e '.[classic_control]'
pip install -e '.[box2d]'

# Other Dependencies
git clone https://github.com/udacity/deep-reinforcement-learning.git
cd deep-reinforcement-learning/python
pip install .

# Kernel
python -m ipykernel install --user --name drlnd --display-name "drlnd"