[Insight-users] Re: levenberg marquadet or other newton-like optimizer for single-value costfunction

[Markus Weigert]

Hi Moti,

Levenberg-Marquardt is indeed designed for non-linear least squares
problems (that' s why it uses a multiple valued cost function).

I dont see why you would need levenberg-marquardt
if you are not dealing with this kind of objective function.
There are lots of competetive optimizers available in the framework,
which allow for fast computation and which are well tested for registration
applications. For example, you may use lbfgs or lbfgsb optimizers which are
quasi newton type optimizers and are quite effective for single valued 
objective functions.


Best regards,
Markus


-----------------------------------------------------------------------------------------
-----------------------------------------------------------------------------------------
Forschungszentrum Juelich GmbH
52425 Juelich

Sitz der Gesellschaft: Juelich
Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498
Vorsitzende des Aufsichtsrats: MinDirig'in Baerbel Brumme-Bothe
Geschaeftsfuehrung: Prof. Dr. Achim Bachem (Vorsitzender), Dr. Ulrich Krafft (stellv. 
Vorsitzender)
-----------------------------------------------------------------------------------------
-----------------------------------------------------------------------------------------