This repository provides implementations of core Deep Reinforcement Learning algorithms, targeting students and researchers learning RL concepts. It offers a structured approach to understanding and experimenting with algorithms like DQN, SARSA, Actor-Critic, DDPG, and AlphaZero, with clear code organization for educational purposes.

How It Works

The project organizes each algorithm into a single, self-contained Python file for ease of understanding and execution. It supports both discrete (CartPole) and continuous (Pendulum) environments from the Gym library, allowing users to test algorithms in different state-action spaces. The code is designed for CPU execution, making it accessible for learning without specialized hardware.

Quick Start & Requirements

• Install dependencies: pip install -r requirements.txt
• Example training command: python -u 04_dqn.py --do_train --output_dir output 2>&1 | tee output/log.txt
• Environments: CartPole (discrete), Pendulum (continuous) from Gym library.
• Tested environment: Python 3.7.
• Hardware: CPU-compatible.

Highlighted Details

• Covers a broad spectrum of RL algorithms from foundational concepts to advanced techniques.
• Includes implementations for single-agent and multi-agent systems.
• Features algorithms like DDPG, TD3 for continuous control and AlphaZero for game AI.
• Demonstrates applications in areas like Neural Architecture Search (NAS) and recommender systems.

Maintenance & Community

No specific information on maintainers, community channels, or roadmap is provided in the README.

Licensing & Compatibility

The repository does not explicitly state a license.

Limitations & Caveats

The code is explicitly stated to be for educational purposes. No performance benchmarks or comparisons are provided. The project does not specify compatibility with Python versions other than 3.7, nor does it detail support for specific operating systems beyond what is implied by Python and Gym dependencies.