Re: ppsp and python

[Daniel Elliott]

John,

Thanks for getting back to me.

On Sat, Oct 20, 2012 at 12:09 PM, John Darrington <address@hidden> wrote:
> On Sat, Oct 20, 2012 at 07:05:04AM -0500, Daniel Elliott wrote:
>
>      My question is this: I have some experience with C/C++
>      interoperability with Python through my efforts with OpenCV.  If I
>      were to add something like neural networks, could integration into
>      PPSP involve writing the IO in C with some hooks into Python to reuse
>      my own, extensive code base?
>
>
> Well of course it *could*.  However, the questions which spring to mind
> are:
>         How much effor would it be?

I am not knowledgeable about the "SPSS way" of doing things with
respect to creating new functions but I figured that,  with neural
nets for example, I could provide the training parameters available
and you could tell me what format the input and output data should be.
 From there, writing the interface wouldn't be too painful.

>         Since the PSPP IO has been designed with SPSS compatibility in mind,
>            would your existing NN implementation fit to the way users would
>            expect to interact with one?

My assumption is that PSPP users are more focused on analyzing results
from a returned model than they are interested in the minutiae of
implementation detail.  From this perspective, I think that the best
way to use an neural network from PSPP would be k-fold
cross-correlation or bootstrap cross-validation which are described in
chapter 6 of Empirical Methods for Artificial Intelligence by Paul
Cohen.  This would shield the user from as many of the issues in model
selection as possible.  It would be good if the users could specify
stuff like the number of layers and the number of nodes in each layer
and the type of activation functions to use or some subset of these
items.  Sadly, the approach to machine learning algorithms is pretty
undisciplined.

> But I'd certainly be interested in any new and innovative  ways of using PSPP.
> Maybe we could add a contrib section.

Again, I am very far from being a competent statistician, but would
enjoy the opportunity to provide some tools to PSPP.  My abilities are
primarily in things like logistic regression, mixtures of Gaussians,
PCA, and neural networks for classification and prediction.  I also do
reinforcement learning but I doubt that is of any use.

Let me know what your thoughts are.

- dan elliott