Машинне навчання для інференції стратегій в іграх

Abstract

This paper considers the problem of inferring fixed opponent strategies in repeated two-player games using artificial intelligence models. In many multi-agent environments, agents interact repeatedly while having limited information about the opponent’s behaviour. Identifying an opponent strategy can significantly improve decision-making. In this work, the opponent strategy inference problem is formulated as a classification task based on observed actions sequences. Artificial intelligence models are used to analyse behavioural patterns and determine the most probable strategy of the opponent. The proposed approach allows an agent to adapt its own behaviour according to the detected strategy. Such methods can be useful in game artificial intelligence and multi-agent systems