Re: [Help-glpk] typo in mip gap formula on wikibook page

[Michael Hennebry]

On Sat, 27 Apr 2013, Heinrich Schuchardt wrote:

> the current definition of the MIP gap is compatible to other optimizers
> like CPLEX.

To the best of of my knowledge,
the terminal output is designed to be read by humans.
It will not be inerpreted by another program.
There is no compatibility issue.

> The advantage of the current definition has the advantage of beeing
> meaningful to the
> end user.
>  
> If the gap is 1 % and the current MIP solution costs me 1 Mio Euro, I
> know that even
> if there is a better solution, it cannot be better by more than 9901
> Euro. Your proposed
> definition would not have such economic significance.

You would not need the 1%.
An optimistic bound is one of the numbers provided on the same line.

On the other hand, were the objective temperature,
the relative gap would change depending on whether
it was expressed in Fahrenheit or Celcius.

My denominator would provide a better indication
of how hard it would be to close the gap.

That said, Andrew has already expressed an unwillingness to change it.

Heinrich Schuchardt quoted Michael Hennebry:
> I'd suggest that a better denominator would be |best_mip - root_lp| That
> would make the formula immune to shifts and scalings. It would also
> ensure that the gap was never greater than 100% .

--
Michael   address@hidden
"On Monday, I'm gonna have to tell my kindergarten class,
whom I teach not to run with scissors,
that my fiance ran me through with a broadsword."  --  Lily