Rapid and brief communication: Kernel ICA: An alternative formulation and its application to face recognition

摘要: This paper formulates independent component analysis (ICA) in the kernel-inducing feature space and develops a two-phase kernel ICA algorithm: whitened principal (KPCA) plus ICA. KPCA spheres data makes structure become as linearly separable possible by virtue of an implicit nonlinear mapping determined kernel. seeks projection directions space, making distribution projected non-gaussian possible. The experiment using subset FERET database indicates that proposed method significantly outperform ICA, PCA terms total recognition rate.