PyRL is a Python Platform for Reinforcement Learning that integrates with many other state of the art RL packages, including the Deep Neural Network based methods.
It is specially designed to include Safe and Survival Reinforcement Learning methods, but also Multi-Armed Bandits, Stochastic Gambling Processes, Dynamic Programming, Classical Planning.
Python scripts and Jupyter notebooks concerning: Pascal's Triangle, Catalan's Triangle, Gambler's Ruin, Decisional Gambling Processes, Multi-Armed Bandits, Survival Bandits, Multi-Armed Gambler, Survival Decision Processes, and Reinforcement Learning
- Constant Agent
- Random Agent
- Policy Gradient
- Proximal Policy Approximation
- Deterministic Policy Gradient
- Q-learning Agent
- Deep Q-Network
- Double DQN
- Dueling DQN
- Actor Critic
- Advantage Actor Critic
- Trust Region Policy Optimization (TRPO)
- Survival Q-learning with budget threshold
- Survival Deep Q-Network
- Pascal's Triangle
- Catalan's Triangle
- Gambler's Ruin
- Survival Bandits