Re: [Help-glpk] 213: No primal feasible solutions

[Andrew Makhorin]

> We have met a problem when solving a MILP model, it is defined as below:

> MIN = sum(ki, i = 1 .. m)

> St:
>  K*y = 0
>  yj >= e
>  0 <= yi <= ki, kiЎК{0,1},1<=i<=m

> K is a n*m coefficient matrix

> y is a m*1 column vector of variables

> The second line in constrains means that the j-th row of y should be a
> positive, so e is boundary positive, which is small enough

> The third line provides relationships bewteen continueous variables y and
> binary variables k

> Well, since m ЎЦ 500, n ЎЦ 30, I don #39;t think it failed because of
> large amount of variables. So are we wrong or it is a bug of glpk?

Please provide more detailed information about the failure. Why do you
think there is a bug?

> the details of linear model is in the attachment.

Could you provide your model written in mps or cplex lp format?