RBF神经网络的混合微粒群学习算法

Hybrid Cooperative Particle Swarm Optimization Learning Algorithm for RBF Neural Network

在分析RBF神经网络的结构特点基础上,定义一个布尔向量L作为网络的结构参数,与原来RBF神经网络的隐节点参数集一起构成了新的RBF网络隐节点参数集{c,,σL},并给出了一个新的RBF网络输入输出关系表达式;采用一种混合协同微粒群算法同时对RBF网络拓扑结构和隐层节点参数进行优化设计,并将输出线性参数集分离后采用最小二乘法进行优化设计,简化了优化空间,加速了算法的收敛速度。

According to the characteristic of RBF neural network, a boolean variable L is defined as the network structural parameter and combined with the original network parameter set to form a new parameter set { c, σ, L } , and a new expression relating input and output of RBF neural network is obtained. Then a hybrid cooperative particle swarm optimization is proposed to optimize the parameter set { c, σ, L } , and output linear parameter set { ω} is separated to optimize using LMS. In the proposed method, the dimension of operating space is reduced and the algorithm convergent is accelerated by the hybrid PSO.