Using of Machine Learning Algorithms without Preliminary Training in Unconstant Game Systems

Abstract

This article contains a comparative analysis of the effectiveness of machine learning algorithms in finding optimal strategy by result in a competitive environment without the possibility of prior training using an example game based on the dilemma of the prisoner. Results of using of adaptive algorithms in comparison with constant strategies are considered from the point of view of game design. Based on the data obtained, the article offers a set of approaches for implementing an adaptive gaming environment as an alternative to the decision trees which are often used in videogame programming.