FURI | Spring 2019

Various Artificial Intelligence Approaches Demonstrated by Chess

The main goal of this research project was to compare the effectiveness of various artificial intelligence (AI) engines in a measurable environment such as chess. In order to train the deep neural network (DNN)  an algorithm to make random movements (RM) was developed. An initial data set consisting of 20,000 games was trimmed to 702; selecting only the games that lead to a checkmate. Thus, demonstrating 3.5% of games with two RM players result in a checkmate. Constructing the DNN involved research into the advantages and disadvantages of various AI methods. Due to the vast chess board possibilities; understanding limitations of various approaches has proven to be key in this research.