论文部分内容阅读
基于人工免疫聚类机制和免疫进化算法,提出了一种新型的设计RBF网络的混合算法。该方法利用人工免疫聚类机制,根据输入数据集合自适应地确定RBF网络核函数的数量及其中心的初始位置。采用免疫进化算法训练RBF网络,进一步缩小了标准进化算法搜索空间的范围,提高了算法的收敛速度。计算机仿真表明,这种RBF网络结构精简并具有较强的泛化能力。
Based on artificial immune clustering mechanism and immune evolutionary algorithm, a new hybrid algorithm for designing RBF network is proposed. The method uses artificial immune clustering mechanism to adaptively determine the number of RBF network kernel functions and the initial position of its center according to the input data set. The immune evolutionary algorithm is used to train the RBF network, which further narrows the search space of the standard evolutionary algorithm and improves the convergence speed of the algorithm. Computer simulation shows that this RBF network structure is streamlined and has strong generalization ability.