Condition number and backward error for the generalized singular value decomposition (Q2706259)

The present paper is a continuation of the author's work [Math. Numer. Sin. 4, 229-232 (1982; Zbl 0559.15013), SIAM J. Numer. Anal. 20, 611-625 (1983; Zbl 0563.65024), Numer. Math. 79, No. 4, 615-641 (1998; Zbl 0907.65038)]. NEWLINENEWLINENEWLINEThe author defines certain normwise condition numbers for a finite generalized singular value of the matrix pair \(\{A,B\}\), where \(A\) and \(B\) are matrices having the same number of columns and \((A^\top,B^\top)^\top\) has full column rank. Explicit expressions of the condition numbers for a simple, nonzero generalized singular value are given. Connections between the presented results and known results in the case of ordinary singular value decompositions are discussed. Furthermore, a normwise backward error of \(\{A,B\}\) with respect to an approximate generalized singular value and an associated approximate generalized singular value group is defined. A computable formula of this backward error is given. The theoretical results are illustrated by two numerical examples.