lhc0512 / RL-framework

Implementation of Deep Reinforcement Learning Algorithms.

RL-framework

Overview

Each algorithm consists of three parts:

• algorithm

• config file (.yaml)

• main function

No inheritance is used, making the algorithm clearer.

Trade-off between performance and readability.

List of Implemented Algorithms

Policy Gradient

• Reinforce

Actor Critic

• Actor Critic (AC)

• Adavantage Actor Critic (A2C)

• Asynchronous Advantage Actor Critic (A3C)

• Generalized Advantage Estimation (GAE)

• Proximal Policy Optimization (PPO)

Value Based

• Deep Q Network (DQN)
• Double Deep Q Network (DoubleDQN)
• Dueling Double Deep Q Network (DuelingDoubleDQN)

Requirements

• Python

• Numpy

• Gym

• Pytorch

• Matplotlib