[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
## Re: [Help-glpk] Two capabilities questions

**From**: |
Andrew Makhorin |

**Subject**: |
Re: [Help-glpk] Two capabilities questions |

**Date**: |
Thu, 20 Dec 2001 17:19:13 +0300 |

>*1) Can GLPK solve problems where the solution vector has bounded but*
>* negative coefficients? The problem I am interested in solving has*
>* coefficients in the range [-1.0, 1.0]. (I am aware that I can fiddle*
>* the problem to require coefficients >= 0 but this further complicates*
>* matters).*
If you mean bounds of structural variables, they can be of any sign.
So, you can set bounds that you need using API like follows:
glp_set_col_bnds(lp, j, 'D', -1.0, +1.0);
or specify them in MPS file in a usual way. For details see "GLPK User's
Guide" included in the distribution.
>*2) Is is possible to feed a known but sub-optimal initial feasible*
>* solution into GLPK and have it optimise that? The problem I am*
>* working on has very tight constraints and many LP solvers have*
>* trouble finding an initial feasible solution.*
>
>*3) Following on from 2), what about an initial solution which is close*
>* to being in the feasible set but isn't.*
You can use API routines glp_simplex1() and glp_simplex2(), which allow
you to solve LP using an initial basis explicitly specified in LPI. For
details see the text file 'newapi.txt' included in the package.