Online Double Oracle (ODO) is a new learning algorithm for two-player zero-sum games where the number of pure strategies is huge or even infinite. Specifically, we combine no-regret analysis from online learning with double oracle methods from game theory. ODO achieves the regret bound
with k being the size of the effective strategy set which is linearly dependent on the support size of Nash equilibrium.
Le Cong Dinh, Yaodong Yang, Zheng Tian, Nicolas Perez Nieves, Oliver Slumbers, David Henry Mguni, Haitham Bou Ammar, Jun Wang (2021) Online Double Oracle https://arxiv.org/pdf/2103.07780.pdf
The code on this repository can be run by cloning the repository
git clone https://github.com/npvoid/OnlineDoubleOracle.gitThe only packages required to run it are scipy, numpy and matpltolib
You can run Leduc Poker by executing
python3 main_leduc.pyYou can run Kuhn Poker by executing
python3 main_kuhn.py
