Is it possible to use SVM to learn a training sample with an input of "Feature Matrix" rather than a "Feature Vector"?

Question

Is it possible to use SVM to learn a training sample with an input of "Feature Matrix" rather than a "Feature Vector" ? I need to classify XML documents by representing each document as a Feature Matrix. Typically, a feature vector is used to train SVM for text classification. However, representing XML documents as feature vectors could lead to structural information loss!

Thanks in advance!

score 1 · Accepted Answer · answered Feb 20 '13 at 06:46

1

Standard practice is to create "long vectors" by "rasterizing" the matrix.

Ultimately, SVMs resolve into lines or hyperplanes, not polygons.

answered Feb 20 '13 at 06:46

Srikant Krishna

881
6
11

Sounds reasonable! Thanks :) – IssamLaradji Feb 20 '13 at 06:55

Is it possible to use SVM to learn a training sample with an input of "Feature Matrix" rather than a "Feature Vector"?

1 Answers1