does matlab 2015 support svm multiclass??

1 vue (au cours des 30 derniers jours)
Sahar abdalah
Sahar abdalah le 22 Mai 2015
Commenté : Walter Roberson le 23 Mai 2015
I use matlab 2013b for programmation but this version does not support svm multiclass. my question there is another version that support svm for multi classification?

Réponse acceptée

Walter Roberson
Walter Roberson le 22 Mai 2015
Modifié(e) : Walter Roberson le 22 Mai 2015
Not directly. There is ClassificationECOC
"ClassificationECOC is an error-correcting output codes (ECOC) classifier for multiclass learning by reduction to multiple, binary classifiers such as support vector machines (SVMs). Train a ClassificationECOC classifier using fitcecoc and the training data."
Real multiclass SVM is really slow and needs a lot of memory, and is considered "feasible" (under a loose sense of the word) for at most 3 classes, and "possible if you are willing to wait several days" for 4 classes, but a matter for supercomputer clusters for 5 or more classes.
  2 commentaires
Sahar abdalah
Sahar abdalah le 22 Mai 2015
thank you for your answer but this type of classification (ClassificationECOC)can handle a large number of classes ?? because I work on a database which contains 1000 classes and I want to know if it is possible to transform the output of SVMs into probability?
Walter Roberson
Walter Roberson le 23 Mai 2015
It looks to me as if it can handle a large number of classes, but the computation is going to involve a decision matrix which is potentially large; see <http://www.mathworks.com/help/stats/classificationecoc-class.html#bue4w15 Coding Design" for how many rows the matrix would have.
I have not investigated how to interpret the various kinds of decision outputs.

Connectez-vous pour commenter.

Plus de réponses (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by