[Insight-users] LBFGS vs LBFGSB

[Luis Ibanez]

Hi Martin,

Thanks for pointing this out.

You are right, the comment in the Example (and therefore in the
ITK Software Guide) is not correct.

The actual reason why the LBFGSB optimizer was used here instead
of the LBFGS optimizer is that several users reported having
trouble with the example when they provided as input the same image
as the Fixed image and Moving image.  In other words, when they
started the registration process by being already at the optimal
value.

It seems that the that implementation of LBFGS in VNL does not
consider that possibility that the optimizer may be started in
the location of the optimal value in the parametric space.


The comment has now been corrected in the CVS version of the
example. In this way it will be fixed for the next edition of
the ITK Software Guide.


Please let us know if you find any other problems,


    Thanks



       Luis



=======================
Martin Urschler wrote:
> hi,
> 
> can anybody explain me the difference between the two limited memory 
> BFGS optimization methods which are implemented in the toolkit?
> itkLBFGSOptimizer
> itkLBFGSBOptimizer
> 
> I understand that the LBFGSB version lets the user specify constraints 
> on the unknowns, however the deformable bspline registration example 
> DeformableRegistration7 argues that "the LBFGSB algorithm is more 
> appropriate for performing optimization in a parametric spaces of higher 
> dimensions"
> I don't see why this is true, IMHO both algorithms are useful for 
> parameter spaces of higher dimensions, since both should be implemented 
> in a "limited-memory" way (that's what the L is for, right?)
> So, how to choose between them? What I find especially interesting in 
> DeformableRegistration7 is that the bounds for LBFGSB are set up in a 
> way that they will be ignored during the optimization, therefore 
> essentially reducing LBFGSB to an LBFGS (with some additional storage 
> requirement due to the more complicated implementation). I'm a little 
> confused with this issue.
> 
> regards,
> Martin
>