[Couenne] integer infeasibility

[Pietro Belotti]

Francois,

I think the current code (both in stable/0.3 and trunk), submitted a few 
days ago, does the same by projecting the LP point back into [l,u] and 
computing the infeasibility.

Pietro

--
Pietro Belotti
Dept. of Mathematical Sciences
Clemson University
email: pbelott at clemson.edu
phone: 864-656-6765
web:   myweb.clemson.edu/~pbelott

On Thu, 26 May 2011, Francois Margot wrote:

> Hi:
>
> I think that one should modify slightly the code in infeasibility.cpp 
> (both trunk and stable) around line 101 by returning 0 for the integer 
> infeasibility if the value is out of bounds. Without this, infinite 
> loops might occur, as the LP solver might return a value slightly out of 
> bounds, by more than the integer tolerance. I would suggest:
>
> if (reference_ -> isInteger ()) {
>  double val = info->solution_[index];
>  if((val >= info->lower_[index]) && (val <= info->upper_[index])) {
>    CouNumber intinfeas = intInfeasibility (info -> solution_ [index]);
>    if ((intinfeas > info->integerTolerance_) && (intinfeas > retval)) {
>      retval = intinfeas;
>      }
>    }
>  }
>
> Francois
>
>
>
> _______________________________________________
> Couenne mailing list
> Couenne at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/couenne
>