CMStatistics 2015: Start Registration
View Submission - CMStatistics
B0733
Title: Canonical correlation analysis for multivariate functional data Authors:  Michio Yamamoto - Kyoto University (Japan) [presenting]
Yoshikazu Terada - National Institute of Information and Communications Technology (Japan)
Abstract: The generalized functional canonical correlation analysis (GFCCA) is the extension of functional canonical correlation analysis from pairs of random functions to the case where a data sample consists of multiple square-integrable stochastic processes more than two. Since the dimension of a function is typically infinite, GFCCA may not be meaningful. To address this issue, sufficient conditions under which GFCCA has a meaningful solution are derived. Furthermore, as with the classical generalized canonical correlation analysis for multivariate data, another formulation of GFCCA based on homogeneous analysis is discovered. The equivalence between the two different formulations is provided, and it enables researchers to use GFCCA for various purposes.