No. 942

Title:

Consistency of Kernel Canonical Correlation Analysis

Author(s):

Fukumizu, Kenji (The Institute of Statistical Mathematics);
Bach, Francis R. (Centre de Morphologie Mathématique, Ecole des Mines de Paris);
Gretton, Arthur (Max Planck Institute for Biological Cybernetics)

Key words:

kenrel CCA; canonical correlation analysis; kernel; consistency; regularization; reproducing kernel Hilbert space

Abstract:

    While the kernel CCA has been applied in many problems, the convergence of the estimated function with finite sample to the true function has not been established yet. This paper gives a mathematical proof of the statistical convergence of kernel CCA and a related method (NOCCO) to provide theoretical justification for the methods. The result gives also a sufficient condition on the regularization coefficient in the methods to ensure convergence.