论文部分内容阅读
为提高网络故障诊断的速度和正确率,提出一种用高斯人工免疫系统(GAIS)来优化BP神经网络权值的方法。GAIS采用概率模型替代传统的变异和克隆操作,是一种分布估计算法。由于高斯网络能准确描述变量之间的联系,避免破坏较优解(构造模块),故此概率模型引用高斯网络。GAIS结合高斯网络和人工免疫系统(AIS)的优点,提高寻优的收敛速度。UCI数据集和网络实测数据集验证了GAIS-BP网络比GA-BP网络收敛速度更快,正确率更高。
In order to improve the speed and correctness of network fault diagnosis, a method of optimizing the weight of BP neural network by Gaussian artificial immune system (GAIS) is proposed. GAIS uses a probabilistic model to replace traditional mutation and cloning operations and is a distribution estimation algorithm. Since Gaussian network can accurately describe the relationship between variables and avoid destroying the optimal solution (constructing modules), the probability model refers to Gaussian network. GAIS combines the advantages of Gaussian network and artificial immune system (AIS) to improve the convergence speed of optimization. UCI dataset and network dataset verify that GAIS-BP network has faster convergence rate and higher correct rate than GA-BP network.