The original RFE article
I. Guyon, J. Weston, S. Barnhill and V. Vapnik.
"Gene Selection for Cancer Classification using Support Vector Machines."
Machine Learning, 2002,
http://www.kyb.tuebingen.mpg.de/bs/people/weston/GENESEL.PDF
appears to describe (Section VI.3, page 30 in the pdf-file) how to
generalize the SVM-RFE to the non-linear case and other kernel methods.
Cheers,
- Kari
Wolfgang Hartmann wrote:
> Hi,
> Only for linear kernel you can compute the nvar(+1) coefficients of
> A linear plane corresponding to the nvar features (plus intercept).
> For nonlinear kernel the weights correspond to the observations and
> Are of course no indicator for the importance of features, have nothing
> To do with the features. So if for linear kernel the coefficients
> of the plane may have some indication of the features of vars (provided
> there is aproximately independece) for the nonlinear case you need
> Other ways to get an estimate how much your fit is reduced (increased)
> When you drop (add) one of the features of the data set.
> HTH, Wolfgang
>
> -----Original Message-----
> From: The Support Vector Machine discussion list [mailto:[log in to unmask]] On Behalf Of Hsiung, Chang (DPAN)
> Sent: Thursday, May 25, 2006 1:55 PM
> To: [log in to unmask]
> Subject: Re: question about SVM-recursive feature elimination
>
> To my knowledge you can not calculate the contributions of each feature
> (variable) in Nonlinear kernel and that is why you can not rank them.
>
> Chang
>
> -----Original Message-----
> From: The Support Vector Machine discussion list [mailto:[log in to unmask]] On Behalf Of samir alwazzi
> Sent: Thursday, May 25, 2006 9:35 AM
> To: [log in to unmask]
> Subject: Re: question about SVM-recursive feature elimination
>
> Hello
> To answer the question, 1st, I have never used SVM-RFE and I just guess what it is about. Also I do not know which piece of software you are using.
>
> Nevertheless I do not see any reason why features could not be ranked when using nonlinear kernels. A simple solution might be to try this case with the software and see if the results are consistent.
>
> Dr.Alwazzi
>
>
>
>
>
>
>> From: Monika Ray <[log in to unmask]>
>> Reply-To: The Support Vector Machine discussion list
>
>> <[log in to unmask]>
>> To: [log in to unmask]
>> Subject: question about SVM-recursive feature elimination
>> Date: Tue, 23 May 2006 21:15:06 -0500
>>
>> Hello,
>>
>> I know that one can use SVM-RFE with nonlinear kernels for feature
>
>> extraction. In the linear kernel case, features are ranked by weight
>
>> and features with lower weights are eliminated. Does the same thing
>
>> happen with nonlinear kernels...such as Radial basis function kernel?
>>
>> Thank You.
>>
>> Sincerely,
>> Monika Ray
>>
>> ***********************************************************************
>> The sweetest songs are those that tell of our saddest thought...
>>
>> Computational Intelligence Centre, Washington University St. louis, MO
>> **********************************************************************
>
> _________________________________________________________________
> FREE pop-up blocking with the new MSN Toolbar - get it now!
>
> http://toolbar.msn.click-url.com/go/onm00200415ave/direct/01/
>
> ______________________________________________________________________
> This email has been scanned by the MessageLabs Email Security System.
> For more information please visit http://www.messagelabs.com/email ______________________________________________________________________
>
> ******************************************
> The information contained in, or attached to, this e-mail, may contain confidential information and is intended solely for the use of the individual or entity to whom they are addressed and may be subject to legal privilege. If you have received this e-mail in error you should notify the sender immediately by reply e-mail, delete the message from your system and notify your system manager. Please do not copy it for any purpose, or disclose its contents to any other person. The views or opinions presented in this e-mail are solely those of the author and do not necessarily represent those of the company. The recipient should check this e-mail and any attachments for the presence of viruses. The company accepts no liability for any damage caused, directly or indirectly, by any virus transmitted in this email.
> ******************************************
|