2012/2/16 Mathieu Blondel <[email protected]>:
> On Thu, Feb 16, 2012 at 3:43 PM, David Warde-Farley
> <[email protected]> wrote:
>
>> I would stress that such an implementation would have to be extremely well 
>> tested and checked against libsvm/liblinear in all of the relevant cases 
>> (obviously, can't check against a float32 case if liblinear doesn't support 
>> float32, etc.)
>
> I believe a Cython port is not that risky if the programmer is well
> organized. I would focus on one solver at a time and would write a
> battery of tests to check the reimplementation against the binding. We
> should be able to reuse (one-vs-rest) or factor out (platt's algorithm
> for probabilistic output) some code too.

True.

>> Moving away from wrappers has one social disadvantage: street cred. NumPy 
>> and SciPy linear algebra routines are trusted because they are, in most 
>> cases, wrappers around battle-hardened Fortran routines that are as old as 
>> the hills. libsvm and liblinear have reputations that a hypothetical 
>> reimplementation effort does not. While SKL is accumulating quite a positive 
>> reputation, a flaky SVM rewrite would be a quick way to undo that.
>
> The reimplementation can be advertised as a "faithful rewrite in
> Cython". This has been done in Java in the past.

Agreed but personally I would rather see such a GSoC student spend
time on writing a cython version of the LaSVM algorithm which looks
much more scalable than libsvm:

  http://leon.bottou.org/projects/lasvm

Although solving for approximate solution.

-- 
Olivier
http://twitter.com/ogrisel - http://github.com/ogrisel

------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing 
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to