A Parameter Selection Optimization Algorithm for Kernel Principal Components Regression

摘要

Kernel principal component analysis can extract the nonlinear features of the data, but the performance is great impact by the parameter of kernel function. This paper presents a kernel parameters optimization method which based on a piecewise binary encoding. The experimental results are very good by using the approach to optimize the kernel parameters in the case of unknown the distribution of the data, which indicating the effectiveness of our method.