Hello colleague, On Thursday, July 17, 2014 6:25:27 AM UTC+2, Stefan Karpinski wrote: > > It's a bit of numerical computing lore that inv is bad – both for > performance and for numerical accuracy. It turns out it may not be so bad > <http://arxiv.org/pdf/1201.6035v1.pdf> after all, but everyone is still > kind of wary of it and there are often better ways to solve problems where > inv would be the naive way to do it. >
i scanned the paper and follows the fashion that people proof numerical assumptions by running matlab experiments ... so i cannot take this seriously. The paper has a point, that if you have random matrix input the method to get Ax = b is -using recent FP implementation- not so influential, by missing the point, that in random input you are not really interested in the output... For real world problems in evaluating ODEs or PDEs where you are especially interested in the result and in ill-conditioned systems, because the impact to the real world would have a price tag (at least, or fatal consequences), i still prefer the conservative solution (of factorisation) which btw (cite from paper): Computing the inverse requires more arithmetic operations than computing an LU factorization.
