Re: [Help-glpk] mipgap accuracy and distance from optimal result

[Yaron Kretchmer]

Great- thanks Andrew.

Kretch

On Mon, Apr 27, 2009 at 10:25 AM, Andrew Makhorin <address@hidden> wrote:

> I #39;m curious about the results glpsol displays during integer

> optimization- What is the relationship (if any) between the percentage
> displayed, and the distance of the result from the optimal.
> So, for instance, if during MIP optimization I see this line

>  1285: mip =  1.825885747e+000 >=  1.667684907e+000   8.7% (74; 1728)

>   Is there a relationship between the 8.7% number and the distance
> between the current best solution(1.825...) and the optimum?

The percentage says that the current objective value 1.825885747e+000
cannot decrease (or increase, in case of maximization) more than on
8.7% of 1.825885747e+000 ~= 0.159. This means the same that the optimal
solution cannot be better than 1.667684907e+000, but expressed in %%.