Thanks Gilles,
I think that the team is doing a fairly good job on the new API, but it is not
yet in a state that I can use,
Further, I have not found what I need in version 3.2 namely
* an implementation of simple gradient descent method
* and implementation of multivariate Newton method
* implementation of other variant of Newton (Quasi-Newton, BFGS etc.)
* And utility to compute gradient and hessian using finite difference.
What I mean is a sort of multivariate version of
FiniteDifferencesDifferentiator that would work to compute gradient ans hessian
for any numerically differentiable function (and not necessarilly one that can
be described using the derivativeStructure API)
In fact, I am facing an optimisation problem for a familly of diffentiables
functions that are a bit complex to evaluate and I want to find the more
efficient method.
I think that I will keep going on my own until the new optim API settle. then I
may have some new optimfeature to propose.
yours truly
François Laferrière
________________________________
De : Gilles <[email protected]>
À : [email protected]
Envoyé le : Mardi 21 mai 2013 0h09
Objet : Re: math : optim package documentation
Hi.
>
> I am, I am a seasonned java developper and I am developping some
> personal research project in DSP.
>
> I need to assess different optim methods for a familly of numerically
> differntiable functions in R^n.
> So far I implemented a few of them on my own (gradient descent,
> Newton, quasi-Newton).
>
> I just discovered that commons-maths 3.2 provide a framework for
> multivariables function. So I would be very happy to use the existing
> optim classes as is, and also, as a base for form own development.
>
> Unfortunately, neither the online dev guide, nor the the javadoc is
> sufficient to use, for instance GradientMultivariateOptimizer : it
> has
> no specific javadoc for the key methods optimise and
> parseOptimisationData.
>
> Is there some documentation somewher to help me?
The developer's guide needs updating.
But in that page:
http://commons.apache.org/proper/commons-math/javadocs/api-3.2/org/apache/commons/math3/optim/nonlinear/scalar/GradientMultivariateOptimizer.html
you can follow the link to the parent class's documentation
in order to find out all the parameters used by the class.
However, before you start writing code that rely on this API,
be sure to read that recent thread:
http://markmail.org/thread/3fjvucyz7rax4cyi
What is your opinion on the proposed change?
Regards,
Gilles
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
