Gathers machine learning and deep learning models for Reinforcement Learning.
All agents will be trained on RGB frames or feature values native from the game engine.
Right now I got machine and time limitation, I stopped working on this repository for a while, I promise I will get back.
ES == Evolution Strategies
DL == Deep Learning
How to read the file, game-folder/algorithm/{features, frames}_{DL, ES}.py
This repository will be update overtime.
Do not try any of these code yet, I never tested it before, still in development.
- Reward based {ES}
- Policy gradient {ES, DL}
- Q-learning {ES, DL}
- Double Q-learning {ES, DL}
- Recurrent-Q-learning {DL}
- Double Recurrent-Q-learning {DL}
- Dueling Q-learning {DL}
- Dueling Recurrent-Q-learning {DL}
- Double Dueling Q-learning {DL}
- Double Dueling Recurrent-Q-learning {DL}
- Actor-Critic {DL}
- Actor-Critic Dueling {DL}
- Actor-Critic Recurrent {DL}
- Actor-Critic Dueling Recurrent {DL}
- Async Q-learning {DL}
- Flappy bird
- level 1 mario
- pong
- Catcher
- Pixelcopter
- Raycast Maze
- Snake
- Water World
- Dooms