脉冲神经网络算法及其在扑克游戏中的应用  

Algorithm of spiking neural network and its application in poker games

作　　者：董丽亚 何虎[1] 王麒淋 杨旭[2] DONG Li-ya;HE Hu;WANG Qi-lin;YANG Xu(Department of Microelectronics and Nanoelectronics,Tsinghua University,Beijing 100084,China;School of Software,Beijing Institute of Technology,Beijing 100084,China)

机构地区：[1]清华大学微电子与纳电子学系,北京100084 [2]北京理工大学软件学院,北京100084

出　　处：《计算机工程与设计》2021年第9期2462-2471,共10页Computer Engineering and Design

基　　金：国家重点研发基金项目(2016YFA0201804)。

摘　　要：随着人工智能的发展,目前主流的神经网络面临着计算量大、功耗高、智能化程度低等问题。为解决以上问题,根据人脑的特性,提出具有普适性的多层脉冲神经网络结构,利用生物学的因果律提出脉冲神经网络算法。通过控制“引导”神经元的激活时间间接调整目标权值,将算法应用在扑克游戏中,使扑克机器人能够学习一个人的打牌能力,实现拟人化程度为85%,验证了算法的可行性,同时表明脉冲神经网络具有强智能性。With the development of artificial intelligence,the current mainstream neural networks are faced with the problems of large computation,high power consumption and low intelligence.To solve the above problems,according to the characteristics of human brain,a universal multi-layer pulsed neural network structure was proposed,and a pulsed neural network algorithm based on the causal law of biology was proposed.The target weight was indirectly adjusted by controlling the activation time of guiding neurons,and the algorithm was applied to poker games,so that the poker robot learned a person’s playing ability.The degree of anthropomorphism is 85%,which verifies the feasibility of the algorithm and shows that the pulsed neural network has strong intelligence.

关 键 词：人工智能 脉冲神经网络 网络结构 学习算法 监督学习 斗地主 突触可塑性规则 赫布规则 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]