|
From: | Ali Baharev |
Subject: | Re: [Help-glpk] control parameters and round-off error |
Date: | Tue, 15 Jan 2008 11:44:51 +0100 |
Thanks for the tips. There was an LP problem where 2.0 <= x_j <= 3.0 and solving min x_j resulted 1.99995. I have lost this example and i do not remember the numbers but it was something like that. It would be difficult to explain the inconsistency with the current one. It is OK if the result is inaccurate, however i would require an error estimate (saying that e.g. the third digit may be incorrect). Computing dot products very precisely is possible using accumulators of C-XSC: http://xsc.de Ali
[Prev in Thread] | Current Thread | [Next in Thread] |