Incremental norm estimation for dense and sparse matrices (Q698566)

It is shown how an incremental norm estimator can be developed for a triangular factor. The suitability of the scheme for both dense and sparse matrices is pointed out due to the fact that one can use approximate singular vectors both from the left and the right hand side. If the explicit inverse of a triangular factor is available, as in the case of an implicit LU factorization, the authors relate the approach to incremental condition estimation. To demonstrate the efficiency of the approach, results are given on some standard test examples for both sparse and dense matrices.