> I have a general question regarding LPs. A LP should have one solution > and so a basis which has normal and dual information. Via changing the > LP (Adding one variable for some inequalities, changing the objective > function and changing the limits of the inequalities a little bit, you > consider a second LP. The dual inequalities from the basis still have > the property that the indicated solution is quite near the new optimal > solution and also 0 is in the dual solution space of the second > problem. > > But the normal information from the basis is not applicable anymore. Is > it possible to work with the old basis in general at LPs and > specifically at glpk?
Glpk simplex solver always starts the search from the current basis specified in lpx object. If you did not remove active constraints and/or basic variables since the last call to lpx_simplex, the basis remains valid that allows the solver performing re-optimization. In particular, glpk b&b solver uses such feature. Andrew Makhorin _______________________________________________ Help-glpk mailing list Help-glpk@gnu.org http://lists.gnu.org/mailman/listinfo/help-glpk