Artificial Intelligence has always had a long history with games. This project seeks to improve our understanding of popular card games played in Casinos using Reinforcement Learning and Deep Learning. The project aims to achieve so, to expand our knowledge of the strategies that are used in Blackjack and Texas Hold’em Poker. Ultimately, helping us understand how we can maximize our profit strategies, why the house always comes out on top and how machines and algorithms can perform complex strategic decisions that surpass human capabilities. The first game the project investigates is Blackjack. The aim is to study and explore various strategies learned by various reinforcement learning algorithms namely, Temporal Difference Learning, Monte Carlo methods, Deep Q Network, and its variants, to improve our understanding of the optimal strategies involved in order to maximize our overall earnings. The next game the paper will tackle, Texas Hold’em Poker, would take our understanding of the algorithms implemented previously further, to build a smart Poker AI, and perform an analysis to understand how to learns against different kinds of Poker players to improve our understanding of machines that can perform complex strategic decisions.
This document contains the introduction to the previous topic, about DeepMind's paper, my idea and the FYP schedule.
This document details the new project, containing the background, current progress, my initial findings, and the next step.
This document details the project, containing the background, the model details, final results and discussion.
The Source Code containing all the Algorithms and Benchmarks Implemented.
The Source Code containing the Texas Hold'em Poker Model.