Re: using interior method with branch and cut

[Andrew Makhorin]

On Sun, 2024-11-10 at 02:03 +0000, Misha Aizatulin wrote:
>    Google says mao@ is the maintainer - hope this works!

This works.

> 
> On 23/10/2024 03:13, Misha Aizatulin wrote:
> > 
> >    Hi, gentle ping please.
> > 
> > On 11/06/2024 13:48, Misha Aizatulin wrote:
> > > 
> > >     Apologies if this is documented somewhere, but I wasn't able
> > > to find anything.
> > > 
> > >     My understanding is that branch_and_cut first calls a linear
> > > solver on the relaxed problem, 
> > > and then tries to convert the solution to an integer solution.
> > > 
> > >     Is there a way to choose which solver we use for the relaxed
> > > part? I would like to try the 
> > > interior method.
> > > 
> 

Using the interior-point method to solve LP relaxations of the node
subproblems in B&B/B&C is not practical for many reasons, e.g., IP is
unable to easily restart the search from the optimal solution of the
parent subproblem (which is dual feasible to the current one).

Glpk MIP solver doesn't provide such a feature.