Constraint Optimization by Linear Approximation

[shaberg]

Hi Fortran users!

I am trying to solve an optimization problem where I need to choose 150 variables to maximize an objective function subject to some non-linear constraints. In my program there is one main constraint and all the other ones just require that the choice variables be strictly positive.

Is there a way to take care of the strictly positive constraints without having to add 150 constraints to the code (some kind of a common requirement to keep the choice variables greater or equal to, say 10^-6)?

Also, how is it related to the code provided online at NLopt:
"call nlo_create(opt, NLOPT_LD_MMA, 2)
call nlo_get_lower_bounds(ires, opt, lb)
lb(2) = 0.0
call nlo_set_lower_bounds(ires, opt, lb)
call nlo_set_min_objective(ires, opt, myfunc, 0)"

What is lb(2) here? Where is it coming from?

Thanks for any help with this!
Shahar

 

[mikrom]

I looked here

http://ab-initio.mit.edu/wiki/index.php/NLopt_Fortran_Reference#Bound_constraints

IMO, lb(2) = 0.0 menas, that lower bonds for 2 variables should be equal zero.
In your cas eof 150 variables, I would try to set lb(150) = 0.0.

 

[shaberg]

Thanks.
So if I choose lb(150)=10^-6 then I restrcit each of the choice variables to be as low as 10^-6?

Thanks,
Shahar

 

[mikrom]

I guess but I'm not sure