IPOPT scaling

[Francisco.Patrocinio]

Hi all,

I have a question: IPOPT returns me this:

Well I have initialized the variables with a plausible, but random value. Is there anyway to circumvent the problem in the figure?

[bussieck]

You should find out why IPOPT can't evaluate the input point. Some solvers don't like gradients that are too large. Usually GAMS already complains if you have a truly bad input point that can't be evaluated (at least for the function evaluation and to some extent for the gradient). IPOPT works also with second order information. Perhaps, the error happens when IPOPT tries to evaluate the Hessian...

-Michael