SVM with uncertain labels 0.2

Remi Flamary, Emilie Niaf, Stephane Canu — Tue, 17 Jul 2012 11:06:23 -0000

SVM are efficient discriminative classifiers but they cannot be applied when the learning set consists of both certain labels {-1,1} and uncertain labels represented by a posterior probability estimate (0,1).

We address this problem in our SSP 2011 paper entitled HANDLING UNCERTAINTIES IN SVM CLASSIFICATION. Basically we learn a unique classifier satisfying both classification performances on the certain labels and performs a probabilistic regression on the uncertain labels. Our approach proved efficient in terms of classification performances and probabilistic output compared to a classical Platt estimation.

This package contains our paper, a matlab function that learn from uncertain labels instead of certain ones (usvmclass.m), and 3 test scripts corresponding to the numerical experiments in the paper (test*.m).

mloss.org SVM with uncertain labels

SVM with uncertain labels 0.2