二打一游戏拆牌算法研究  

Research on card splitting algorithm of poker game Two-against-One

作　　者：乔秀明 黄文杰 李淑琴[1,2] QIAO Xiuming;HUANG Wenjie;LI Shuqin(Computer School,Beijing Information Science&Technology University,Beijing 100101,China;Perception and Computational Intelligence Laboratory,Beijing 100101,China)

机构地区：[1]北京信息科技大学计算机学院,北京100101 [2]感知与计算智能联合实验室,北京100101

出　　处：《重庆理工大学学报（自然科学）》2022年第12期170-176,共7页Journal of Chongqing University of Technology：Natural Science

基　　金：北京信息科技大学科技项目(5212010937,KM201911232002,5112011019,5112011041)。

摘　　要：为模拟人类玩家在二打一打牌过程中的拆牌规律,提出了基于深度神经网络的拆牌训练,采用基于深度学习的序列标注方法,使用大量经过标注的实战数据,采用BILSTM-CRF网络构建手牌拆牌识别网络,实现了对二打一初始手牌的拆牌目标。实验结果表明:该模型在原始数据中产生了良好的训练效果,在测试集上的准确率达到89.16%,在与传统的手牌拆分方法进行比较后印证了该方法的可靠性,为进一步对二打一初始手牌的难度评估提供了保障,也为其他非完全信息牌类博弈的手牌拆分方法提供了借鉴。Splitting hand cards of poker game Two-against-One is a preprocessing task for analyzing the winning rate of player hand cards,which affects the subsequent research on the the winning rate.However,because the process of playing includes single out and a combination of multiple cards,card splitting is an indispensable step in the study of hand cards.Aiming at how to simulate human players to split hand cards in Two-against-One,this paper proposes a splitting training based on a deep neural network.Inspired by the sequence labeling method in deep learning,the experiment collects a large number of labeled data from real scenes,and uses BILSTM-CRF network to construct the hand card splitting recognition model,so that it can split initial hand cards in Two-against-One.Experiments shows that the model achieves a good training performance on the original test data with an accuracy of 89.16%in the test set.Through the comparison with the traditional hand card splitting method,the reliability of the model is verified,thus providing a guarantee for the further evaluation of the difficulty of initial hand cards,and also providing valuable experience of hand card splitting methods for other incomplete information card games.

关 键 词：二打一 拆牌算法 深度学习 序列标注 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]