pearl-rl 0.1.0

A reinforcement learning library with explainability features

Pearl RL

A reinforcement learning library with explainability features for AI/ML research and development.

Features

• Reinforcement Learning Agents: Implementation of various RL algorithms including DQN
• Environment Wrappers: Easy integration with Gymnasium environments
• Explainability Methods: Multiple explainability techniques including LIME, SHAP, and LMUT
• Stability Analysis: Tools for analyzing model stability
• Pre-built Environments: Support for Atari games and Lunar Lander

Installation

pip install pearl-rl

Quick Start

from pearl import Pearl
from pearl.agents import SimpleDQN
from pearl.enviroments import GymRLEnv

# Create an environment
env = GymRLEnv("LunarLander-v2")

# Create an agent
agent = SimpleDQN(env.observation_space, env.action_space)

# Create Pearl instance
pearl = Pearl(agent, env)

# Train the agent
pearl.train(episodes=1000)

# Get explanations
explanations = pearl.explain(episode=0)

Documentation

For detailed documentation and examples, please visit our documentation page.

Contributing

We welcome contributions! Please see our contributing guidelines for details.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Citation

If you use Pearl in your research, please cite:

@software{pearl_rl,
  title={Pearl RL: A Reinforcement Learning Library with Explainability Features},
  author={Your Name},
  year={2024},
  url={https://github.com/yourusername/pearl}
}