First and second-order derivatives for CP and INDSCAL

Abstract In this paper we provide the means to analyse the second-order differential structure of optimization functions concerning CANDECOMP/PARAFAC and INDSCAL. Closed-form formulas are given under two types of constraint: unit-length columns or orthonormality of two of the three component matrices. Some numerical problems that might occur during the computation of the Jacobian and Hessian matrices are addressed. The use of these matrices is illustrated in three applications.

[1]  J. Chang,et al.  Analysis of individual differences in multidimensional scaling via an n-way generalization of “Eckart-Young” decomposition , 1970 .

[2]  H. Law.  Research methods for multimode data analysis , 1984 .

[3]  J. Leeuw,et al.  Explicit candecomp/parafac solutions for a contrived 2 × 2 × 2 array of rank three , 1988 .

[4]  P. Paatero.  A weighted non-negative least squares algorithm for three-way ‘PARAFAC’ factor analysis , 1997 .

[5]  Mohammed Bennani Dosse,et al.  The Assumption of Proportional Components when Candecomp is Applied to Symmetric Matrices in the Context of Indscal , 2008, Psychometrika.

[6]  A. Stegeman.  Degeneracy in Candecomp/Parafac and Indscal Explained For Several Three-Sliced Arrays With A Two-Valued Typical Rank , 2007, Psychometrika.

[7]  Richard A. Harshman,et al.  Foundations of the PARAFAC procedure: Models and conditions for an "explanatory" multi-model factor analysis , 1970 .

[8]  A. Stegeman.  Degeneracy in Candecomp/Parafac explained for p × p × 2 arrays of rank p + 1 or higher , 2006 .

[9]  Jorge N. Tendeiro,et al.  Some New Results on Orthogonally Constrained Candecomp , 2011, J. Classif..

[10]  J. Kruskal.  Three-way arrays: rank and uniqueness of trilinear decompositions, with application to arithmetic complexity and statistics , 1977 .

[11]  Henk A. L. Kiers,et al.  Some clarifications of the CANDECOMP algorithm applied to INDSCAL , 1991 .

[12]  J. Berge,et al.  A joint treatment of varimax rotation and the problem of diagonalizing symmetric matrices simultaneously in the least-squares sense , 1984 .

[13]  R. A. Harshman,et al.  Data preprocessing and the extended PARAFAC model , 1984 .

[14]  Henk A. L. Kiers,et al.  A treatment of the Orthomax rotation family in terms of diagonalization, and a re-examination of a singular value approach to Varimax rotation , 1988 .

[15]  J. Magnus,et al.  Matrix Differential Calculus with Applications in Statistics and Econometrics , 1991 .

[16]  P. Paatero.  The Multilinear Engine—A Table-Driven, Least Squares Program for Solving Multilinear Problems, Including the n-Way Parallel Factor Analysis Model , 1999 .