A square root merit function for Canonical Correlation Analysis

Canonical Correlation Analysis (CCA) is a well-known technique in multivariate statistical analysis, which has been widely used in economics, meteorology, and in many modern information processing fields. This paper proposes many dynamical systems for computing canonical correlations and canonical variates. These systems are shown to converge to the actual components rather than to a subspace spanned by these components. Qualitative properties of the proposed systems are analyzed in detail including the limit of solutions as time approaches infinity. Convergence is illustrated by a numerical example.