核独立成分分析在fMRI数据中的应用

介绍了一种全新的功能核磁共振数据（fMRI）分析方法即核独立成分分析方法,它是核方法与典型相关分析方法的综合运用。将Bach和Jordan提出的核独立成分分析方法应用于功能核磁共振成像数据分析,并基于实验所得的时间序列与参考函数的相关系数的大小,比较了该方法与经典的成分分析方法-主成分分析（PCA）方法与快速不动点算法（FastICA）的实验结果。结果表明,对fMRI数据而言,核独立成分分析方法和FastICA算法要优于PCA算法。

Kernel Independent Componet Analysis for fMRI Data

A novel functional magnetic resonance imaging(fMRI)data analysis method is introduced.The proposed method,referred as kernel independent component analysis(Kernel-ICA),combines the strength of the kernel approaches with canonical correlation analysis.For performing Kernel-ICA,an algorithm developed by Bach and Jordan is employed.Based on the analysis of fMRI data,the performance of Kernel-ICA is compared with classical algorithms such as PCA and FastICA.The comparative results are based on correlation and associated timecourses.Numerical results show that both Kernel-ICA and FastICA are outperformed to PCA for analyzing the fMRI data.