On the condition number of the total least squares problem (Q368579)

The authors consider the singular value decomposition (SVD) based computable formulas and bounds estimates for the condition number of the total least squares (TLS) problems. The closed relation between SVD and TLS problems are explored permitting the authors to obtain new closed formula for the TLS condition number using weaker conditions. Furthermore these bounds improve the existing theory and can be applied to large scale TLS problems. The advantage of this new formulation is in the use of only the matrix \([A,b]\) instead of both \(A\) and \([A,b]\) and also the stablishment of both lower and upper bounds that require at most the smallest two singular values. Numerical experiments validate the presented theory showing the improvement in relation to the previous known results.