Similar to Deep CFR and Poker-CNN (see papers below), this is an implementation of the External Sampling CFU (Probing) flavor of Monte-Carlo Counterfactual Regret Minimization, using XGBoost for the model(s). The current tree generation is for Heads-Up Limit Poker.
Disclaimer: These are not 'best' programming practices (it's written in VB after all), it's not very well commented, and it's not very efficient. :) I make no guarantees as it's function or performance (I haven't tested it with a GPU) and I take no responsiblity for how this code is used or misused.
Deep CFR:
https://arxiv.org/pdf/1901.07621.pdf
https://arxiv.org/pdf/1811.00164.pdf
https://github.com/EricSteinberger/Deep-CFR
(Much respect to Eric Steinberger. That kid is only 19.)
Poker-CNN:
https://arxiv.org/pdf/1509.06731.pdf
HoldemHand:
https://www.codeproject.com/Articles/12279/Fast-Texas-Holdem-Hand-Evaluation-and-Analysis
SharpLearning/XGBoost:
https://github.com/mdabros/SharpLearning