Training Classifiers for Tree-Structured Sets of Categories
D. Gutiérrez-González, Manuel Ortega-Moral, M. L. De-Pablo, Jesús Cid‐Sueiro · 2006
In this paper we propose a new method for training classifiers for multi-class problems when classes are not (necessarily) mutually exclusive and may be related by means of a probabilistic tree structure. Our method is based on the definition of a Bayesian model relating network parameters, feature vectors and categories. Learning is stated as a maximum likelihood estimation problem of the classifier parameters. The proposed algorithm is tested on an image retrieval scenario