用遗传算法优化Boltzmann机  

Boltzmann machine optimization based on genetic algorithm

作　　者：陈洁[1] 刘希玉[1] 姚树魁[1] 

机构地区：[1]山东师范大学管理与经济学院,济南250014

出　　处：《计算机工程与应用》2011年第7期46-48,60,共4页Computer Engineering and Applications

基　　金：国家自然科学基金重大项目No.60873058;No.60743010;山东省自然科学基金重大项目(No.Z2007G03);"泰山学者"建设工程专项经费资助项目;山东省教育厅计划项目~~

摘　　要：Boltzmann机是一种应用广泛的随机神经网络。它通过模拟退火算法进行网络学习,能取得一个全局或接近全局最优的最优值;通过期望网络模式和实际学习得到的网络模式比较来调节网络的权值,使网络能尽可能地达到或逼近期望的网络模式。将遗传算法运用到Boltzmann机的网络学习中,在对BM机编码后,通过选择、交叉和变异等遗传操作算子对网络进行训练,调整网络的权值,使适应度函数值大的网络保留下来,最终使网络达到期望的模式。通过实例验证,这是一种简单可行的调节网络权值的方法。Boltzmann machine is widely used in stochastic neural networks.It can obtain the global optimum or a near-optimal solution by network learning based on simulated annealing.Compared the expectations of the network model with the actual learning of the network model to adjust the weights of the network,so that the network can be as much as possible to meet or close to the desired network model.This genetic algorithm is applied to the learning of the Boltzmann machine,which can change the network weights through the competitive selection genetic algorithm optimizing operation,so that the network can achieve the desired pattern.An example proves that this is a simple and practical way to adjust the network weights.

关 键 词：BOLTZMANN机 遗传算法 网络模式 

分 类 号：TP389.1[自动化与计算机技术—计算机系统结构]