 |
    |
Large Margin Local Metric Learning
Julien Bohné1, 2, Yiming Ying3, Stéphane Gentric2, and Massimiliano Pontil1
1University College London, Department of Computer Science, London, UK
julien.bohne@morpho.com
m.pontil@cs.ucl.ac.uk
2Safran Morpho, Issy-les-Moulineaux, France
stephane.gentric@morpho.com
3University of Exceter, Department of Computer Science, Exceter, UK
y.ying@exeter.ac.uk
Abstract. Linear metric learning is a widely used methodology to learn a dissimilarity function from a set of similar/dissimilar example pairs. Using a single metric may be a too restrictive assumption when handling heterogeneous datasets. Recently, local metric learning methods have been introduced to overcome this limitation. However, they are subjects to constraints preventing their usage in many applications. For example, they require knowledge of the class label of the training points. In this paper, we present a novel local metric learning method, which overcomes some limitations of previous approaches. The method first computes a Gaussian Mixture Model from a low dimensional embedding of training data. Then it estimates a set of local metrics by solving a convex optimization problem; finally, a dissimilarity function is obtained by aggregating the local metrics. Our experiments show that the proposed method achieves state-of-the-art results on four datasets.
Keywords: Dissimilarity learning, local metric learning, face recognition, nearest neighbor classification
LNCS 8690, p. 679 ff.
lncs@springer.com
© Springer International Publishing Switzerland 2014