Text Categorization via Ellipsoid Separation
Andriy Kharechko, John Shawe-Taylor, Ralf Herbrich and Thore Graepel
In: Learning Methods for Text Understanding and Mining, 26 - 29 January 2004, Grenoble, France.
We present a new batch learning algorithm for text classification in the vector space of document representations. The algorithm uses ellipsoid separation in the feature space which leads to a semidefinite program. An approximation of the latent semantic feature extraction approach using Gram-Schmidt orthogonalization is used for the feature extraction. Preliminary results demonstrate some potential for the presented approach.