一种并行遗传优化核主元分析算法

针对核主元分析算法参数设置和主元数量选取问题，提出一种并行优化核函数参数和主元数量的改进核主元分析算法。该算法以类别可分性为准则，应用自适应遗传算法同步对核参数和主元数量进行优化，实现了核参数和主元数量的并行选择。将改进后的核主元分析算法应用于柴油机气阀机构典型故障的特征提取中，结果表明：优化核主元分析能有效降低柴油机气阀机构故障特征向量的维数，提高各类样本的聚类效果。

Algorithm for Kernel Principal Component Analysis Based on Parallel Genetic Optimization

Aiming to kernel parameters settings and principal components selection of kernel principal component analysis, an improved KPCA algorithm with parallel optimization of kernel function parameters and principal components number is put forward. Base on class separability criterion, the algorithm applys adaptive genetic algorithm to optimize the kernel parameters and principal components number synchronously, achieving parallel selection of kernel parameters and principal components number. The improved KPCA algorithm is applied to feature extraction of typical faults for diesel valve train. The results show that optimal KPCA can effectively reduce dimensions of fault feature vectors, improve samples clustering effect of various classes and lay a good foundation for classification precision improvement.