This project uses multiple reinforcement learning algorithms to train agents for the holdem texa version of the poker game.
Table of Contents
In this project we explored the basic reinformcent learning arlogirtms (SARSA, Expected SARSA, Q Learning) and some based on neural networks (DQN and its variants) to see how each algorithm would perform in a game of poker. Our objective was to find the best model with whom to play the game.
The models were trained agains the random model which comes with the enviroment, to avoid the possibility of them to learn the patter of their adversary rather than developing a true strategy for the game. Each model has its own logic and learning mechanism on which it bases its behavior.
- SARSA
- Expected SARSA
- Q Learning
- DQN Base
- DQN Target Network
- DQN Target Network and Experience Replay
- DQN All
For more details regarding each agent please refer to the docs »
You need to have python 3 and pip 3 already installed on your machine.
-
Clone the repo
git clone https://github.com/claudia-maria-dudau/RL-Poker.git -
cd into this project
-
Install the environment from github
pip install -e git+https://github.com/dickreuter/neuron_poker#egg=neuron_poker -
Install necessary libraries
pip install -r requirements.txt -
Run project
pythom main.py [option]
Project Link: https://github.com/claudia-maria-dudau/RL-Poker
This project was made by:
- Agha Mara
- Buduroes Bianca
- Dudau Claudia
- Poinarita Diana