scientific article; zbMATH DE number 3982395

From MaRDI portal
Publication:3746786





A generalization of the power method called the power-successive overrelaxation (P-SOR), is derived for the numerical solution of a class of multiparameter eigenvalue problems, i.e. to find a diagonal matrix \[ \Lambda =diag(\lambda_ 1I^{(n_ 1)},...,\lambda_ mI^{(n_ m)})\geq 0 \] and a corresponding real vector \( x=(x^ T_ 1,...,x^ T_ m)^ T,\) \(x_ i:\) \(n_ i\)-dimensional vector \((1\leq i\leq m)\) satisfying the following equations: \[ Ax=\Lambda x,\quad \| x_ i\| =1,\quad i=1,2,...,m, \] where \(A=(A_{ij})\) is a real symmetric block matrix, \(A_{ii}\) are \(n_ i\times n_ i\) positive definte matrices, \(1\leq i\leq m,\| \cdot \|\) denotes the Euclidean norm. The method applied to canonical correlation analysis of several sets of variables. The computational results show that the P-SOR method is more effective than the algorithm of \textit{P. Horst} [Psychometrika 26, 129-149 (1961; Zbl 0099.358)].




Cited in
(21)








This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3746786)