Using solution of one model as upper/lower bound in that model again - Global Optimization

[zar]

Hi everyone,

I working on a global optimization problem, where I want to improve solution search space of the algorithm by using solution of local NLP solver as an upper/lower bound in the relaxed version of that model again. How can I achieve that:
(a) By defining both models in the same file i.e above and below of each other
(b) By defining them in different files

Appreciate your help.

Thanks
ZAR

[bussieck]

Be careful with "helping" an algorithm with bounding scheme (like branch-and-bound) with bounds. For minimization problem only set an upper bound (or even better provide a feasible solution). Never set the the lower bound. (opposite for maximization problems). Setting a "helping" lower bound deprives the algorithm from making good algorithmic choices (e.g. variable and node selection) based on the relaxation (lower bound), since lower bounds of all relaxations are artificially raised to the same lower bound.

-Michael