基于误差模型的自适应鲁棒主成分分析

研究了改善主成分分析(PCA)算法鲁棒性的一种实现途径.通过对误差函数的建 模分析,得到一种改进的目标函数.提出一种新的在线自适应式的鲁棒PCA运算规则.该方 法基于单层线性神经网络(NN)结构,但是权值的训练算法是非线性的.从而在迭代训练中对 "劣点"样本加以适当处理来排除对运算精度和收敛性的影响.

ADAPTIVE ROBUST PRINCIPAL COMPONENT ANALYSIS BASED ON ERROR MODELING

One way to improve the robustness of principal component analysis (PCA) is studied in the paper. A new adaptive algorithm of robust PCA based on the structure of single layer neural network (NN) is developed with modification of the cost function which can be acquired through modeling of the error function. The new nonlinear robust PCA algorithm can reduce the effects of outliers on the accuracy and convergence of the PCA algorithm through proper processing of them.