A Highly-Parameterized Ensemble to Play Gin Rummy
Published in Proceedings of the AAAI Conference on Artificial Intelligence, 2021
This paper presents the design and training of an autonomous Gin Rummy player. The system consists of three core components responsible for decision-making in card drawing, discarding, and game termination. An ensemble-based strategy is employed to improve discard decisions, and a genetic algorithm is used for model tuning. The paper concludes with an analysis of parameter optimization and performance evaluation across three experimental configurations.
Nagai, Masayuki, Kavya Shrivastava, Kien Ta, Steven Bogaerts, and Chad Byers. "A Highly-Parameterized Ensemble to Play Gin Rummy." Proceedings of the AAAI Conference on Artificial Intelligence, vol. 35, no. 17, pp. 15614-15621. 2021.
Download Paper