This repository provides minimal and clean code examples for reinforcement learning algorithms, targeting learners from basic concepts to deep reinforcement learning. It offers easy-to-read, single-file implementations for various algorithms, facilitating understanding and experimentation.

How It Works

The project implements reinforcement learning algorithms in a modular, single-file-per-algorithm approach. This design choice emphasizes clarity and ease of understanding, allowing users to focus on the core logic of each algorithm without navigating complex project structures.

Quick Start & Requirements

• Install: pip install -r requirements.txt
• Prerequisites: Python 3.5, Tensorflow 1.0.0, Keras, numpy, pandas, matplot, pillow, Skimage, h5py.

Highlighted Details

• Covers fundamental RL concepts with Grid World examples (Policy Iteration, Value Iteration, SARSA, Q-Learning).
• Demonstrates deep RL applications on CartPole (DQN, Double DQN, Policy Gradient, A2C, A3C).
• Includes Atari game implementations (Breakout, Pong) using advanced DQN variants and A3C.

Maintenance & Community

Maintained by the RLCode team (Woongwon, Youngmoo, Hyeokreal, Uiryeong, Keon). Open to pull requests and issues.

Licensing & Compatibility

The repository does not explicitly state a license in the provided README.

Limitations & Caveats

The project requires an older version of Python (3.5) and Tensorflow (1.0.0), which may pose compatibility challenges with modern development environments and libraries. The OpenAI GYM Mountain Car example is marked as "[WIP]".