Installation

git clone https://github.com/kkugosu/RL_BASIC.git

Experiment

python executable.py

you can choose belows

• environments
• policy
• hiddenlayer size
• batch size
• memory capacity
• memory reset time
• train time per memory
• learning rate
• eligibility trace
• done penalty
• load previous model or not (not = 0 yes = 1)

Example

result of reward > 500

Experiment environment

• cartpole
• hopper

Requirements

• Gym
• Mujoco
• Python >= 3.8
• Pytorch >= 1.12.0
• Numpy

Easy explained algorithm principle

policy gradient

deep queue network

actor critic

deep deterministic policy gradient

Trust region policy optimization

proximal policy optimization

soft actor critic

REPO

• papers

• cs285

• hui reinforcement learning blog