On Fri, 9 Jul 2004, Shwetha Harish wrote:
> Hi !
>
> I am trying to use SVM for spam filtering.
>
> I have categorized a bunch of emails into spam and non-spam and I have y =
> {-1,+1} and alos xi.
>
> I am unable to understand as to how to calcalute the optimal weight vector
> and the constant b.
>
> I am trying to formula yi(w.xi-b)>=1
>
> Could you please help me.
It is an interesting application. I have implemented SMO i Java, and also
some methods for faster training. This is described in a technical report
by Keerthi et al.:
@TECHREPORT{keerthi99improvements,
author = {S. Keerthi and S. Shevade and C. Bhattacharyya and K. Murthy},
title = {Improvements to Platt's SMO algorithm for SVM classifier
design},
institution = {National University of Singapore},
year = {1999},
text = {S. Keerthi, S. Shevade, C.
Bhattacharyya and K. Murthy. Improvements to Platt's SMO
algorithm for SVM
classifier design. Tech Report, Dept. of CSA, Banglore, India,
1999.},
url = {citeseer.ist.psu.edu/244558.html},
groupsearch = {0},
}
If you want my Java code then just e-mail. There are also good open Java
implementation in the LIBSVM package.
Sincerely,
Rasmus P.
> Thanks in advance
> Shwetha Harish
>
|