Re: [Help-glpk] Preprocessing

[glpk xypron]

Hello Andrew,

> Did you intensively test that transformation?
No, testing is still required. This is just a draft version.

> make sure that all basic solution components (statuses, primal and dual
> values) are recovered correctly. Besides, can it be applied to interior
> point and mip solutions?
If I find two rows:
lb1 <= sum(aj xj) <= ub1
lb2 <= sum(bj xj) <= ub2
such that aj * f = bj for all j with f = const
I use
   row1->row->lb = -DBL_MAX, row1->row->ub = +DBL_MAX;
   npp_free_row(npp, row1->row);
to free the rows,

I create a new row
   row = npp_add_row(npp);
such that
max(lb2, lb1 * f) <= sum(bj xj) <= min(ub2, lb1 *f )

This should not change status, primal, dual values. But this needs
testing.

Best regards

Xypron

-------- Original-Nachricht --------
> Datum: Sat, 16 Oct 2010 01:12:50 +0400
> CC: Kevin Hunter <address@hidden>, Help Glpk <address@hidden>
> Betreff: Re: [Help-glpk] Preprocessing

> Hi Xypron,
> 
> > I implemented an elimination of linear dependent rows in appended
> > glpnpp06.c.
> > 
> > This cuts the solution time for Kevin's problem down from several
> > hours to 9 minutes.
> 
> Great! Did you intensively test that transformation? It is important to
> make sure that all basic solution components (statuses, primal and dual
> values) are recovered correctly. Besides, can it be applied to interior
> point and mip solutions?
> 
> 
> Andrew Makhorin
> 
> 

-- 
Neu: GMX De-Mail - Einfach wie E-Mail, sicher wie ein Brief!  
Jetzt De-Mail-Adresse reservieren: http://portal.gmx.net/de/go/demail