Prev: Output dimensions in L2 M-files S-function block
Next: Non-Normally distributed data
From: Ben on 7 Oct 2009 05:33
Hi, all.
I'm newbie in Matlab. I would like to know whether svmtrain and svmclassify could be used on my multi-dimensional dataset? I know it could handle 2D data well but I'm not sure whether it could classify data with more than 2D correctly?
From: Novae on 7 Oct 2009 06:34
Hi Ben and welcome to Matlab

You can use svmtrain and svmclassify for a multi-dimensional data set. There's no limit to the feature space dimension when you use a svm classifier. However, depending on this dimension and kernel used, training phase can take a long time. Correct classification depends on a lot of things: descriptor quality, kernel, cost, etc... Remember that svm classification using the bio-informatics toolbox only allows binary data to be classified. This means that your data should have only 2 classes. If you have more than 2 classes I would recommend the usage of svmlib library available on the internet. The usage is very similar to the bio-informatics toolbox. Just search for it and have fun.

Igor

"Ben " <weehow911(a)gmail.com> wrote in message <hahn8f$621$1(a)fred.mathworks.com>...
> Hi, all.
> I'm newbie in Matlab. I would like to know whether svmtrain and svmclassify could be used on my multi-dimensional dataset? I know it could handle 2D data well but I'm not sure whether it could classify data with more than 2D correctly?
 | 
Pages: 1
Prev: Output dimensions in L2 M-files S-function block
Next: Non-Normally distributed data