Using CNN for solving two-player zero-sum games. (15th October 2022)

• Record Type: Journal Article
• Authors: Wu, Dawen; Lisser, Abdel
• Is Part Of: Expert systems with applications. Volume 204(2022)
• Publication Date: 2022-10-15
• Subjects: Two-player zero-sum game -- Saddle point -- Convolutional neural network -- Machine learning; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2022.117545
• Languages: English
• ISSNs: 0957-4174

Abstract

Abstract: We study a two-player zero-sum game (matrix game for short) with the objective of finding the saddle point and its value. We develop a novel convolutional neural network (CNN for short) approach to achieve the goal. We propose a complete training pipeline, including a specific CNN model structure to handle varying game sizes, generating training datasets, and model fitting. The experiment results show that our proposed method outperforms the traditional linear programming (LP for short) method and two regret minimization learning algorithms in terms of computational efforts. Highlights: We use a novel CNN method to solve two-player zero-sum games. Concrete training algorithms are proposed to train the CNN model for the games. Our CNN model can handle different game sizes and untrained generation distributions. Our CNN approach shows great potential in terms of computational efficiency.