Deep Kernel Principal Component Analysis for multi-level feature learning

Francesco Tonin, Qinghua Tao, Panagiotis Patrinos, Johan A. K. Suykens

Abstract: Highlights•A deep kernel principal component analysis (DKPCA) framework is proposed.•Forward and backward couplings between the levels are identified.•Theoretical analysis presents error bounds and higher explained variance than KPCA.•Generative DKPCA is introduced for the pre-image problem in deep kernel methods.•DKPCA is competitive in learning informative multi-level disentangled features.