[Ipopt] hessian option

[Sebastian Walter]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Sourav Rakshit schrieb:
> Hi all,
> I am new to ipopt and trying to use it for energy minimization in protein
> structure prediction. Unfortunately, my function is very complex and
> calculating gradient and hessian analytically involves a lot of work ;
> specially hessian calculation is almost impossible for me. I am working
> with ma27 solver...i tried mumps but it fails. Unfortunately, IPOPT is not
> able to converge for my problem. Can I increase the performance of IPOPT
> by suitable options? Till now i have no constraints in my problem and I am
> not providing the hessian.
> sincerely,
> Sourav Rakshit
> 

You could use Automatic Differentiation (AD) to obtain the Hessian and gradient.
If you are lucky, you need only small code modifications to compute the derivatives.
There is a website dedicated to Automatic Differentiation:
      http://www.autodiff.org/
AD is suitable for complex simulations and has distinct advantages over finite differences and
symbolic calculations.

best regards,
Sebastian Walter

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.4-svn0 (GNU/Linux)
Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org

iD8DBQFKCSRD9PBA5IG0h0ARAvjpAJsFAZDAtnADrW7rVf4g5DKRIrJeqACfSqXw
CYKU3arflNJ85eWaBJtu2hk=
=qT14
-----END PGP SIGNATURE-----