> Thanks for the explanation. If I understand > well this is a case where the default scaling > (geom mean + equil scaling but skip if LP is well scaled) > gives bad results.
Yes. You may look at max|aij| and min|aij| before and after scaling; these are displayed on the terminal. > I was really surprised so I sent this > bug report (as geometrically the problem appears > very simple, and also any couple of neighbouring constraints > should gives a very well conditionned matrix). The glpk simplex solver is insufficiently smart to improve the original lp instance. > > Btw when I set the scaling to: > > geometric mean scaling + round scale factors to power of 2 > equilibration scaling > equilibration scaling + round scale factors to power of 2 > > I get an accurate result. > > But not with: > > geom mean + equil scaling + round scale factors to power of 2 > > And as you said no scaling is good too here. So for this particular > constraint matrix, equilibration scaling seems less sensible to this > problem of a tiny value. > GM scaling works well if the problem is badly scaled, but not in the case of tiny coefficients. > PS : The tiny value comes from the fact that the problem was set > in an automatic way from a numerical code. This is a frequent error. The program that generates lp data should make necessary checks and replace tiny coefficients with exact zero. _______________________________________________ Bug-glpk mailing list [email protected] https://lists.gnu.org/mailman/listinfo/bug-glpk
