This article is cited in 1 paper

Part 1. General problems of the intellectual systems theory

Relationship between accuracy of singular value decomposition and distribution of singular values

I. Yu. Drozdov, D. V. Parfenov

MIREA — Russian Technological University, Moscow

Abstract: Using numerical experiments we demonstrated that accuracy of numerical singular value decomposition (SVD) is affected not only by condition number of the matrix, but also by distribution of singular values in the matrix spectrum. We note that widely used relative norm error of SVD result might be insufficient and propose an alternative metric based on root-mean-squared relative error.

Keywords: singular value decomposition, SVD, condition number, matrix spectrum, numerical stability.