Latent-lSVM classification of very high-dimensional and large-scale multi-class datasets

摘要

We propose a new parallel learning algorithm of latent local support vector machines(SVM), called latent-lSVM for effectively classifying very high-dimensional and large-scalemulti-class datasets. The common framework of texts/images classification tasks using theBag-Of-(visual)-Words model for the data representation leads to hard classification problemwith thousands of dimensionsandhundreds of classes.Ourlatent-lSVM algorithmperforms thesecomplex tasks into two main steps. The first one is to use latent Dirichlet allocation for assigningthe datapoint (text/image) to some topics (clusters) with the corresponding probabilities. Thisaims at reducing the number of classes and the number of datapoints in the cluster comparedto the full dataset, followed by the second one: to learn in a parallel way nonlinear SVM modelsto classify data clusters locally. The numerical test results on nine real datasets show that thelatent-lSVM algorithm achieves very high accuracy compared to state-of-the-art algorithms. Anexample of its effectiveness is given with an accuracy of 70.14% obtained in the classification ofBook dataset having 100 000 individuals in 89 821 dimensional input space and 661 classes in11.2minutes using a PC Intel(R) Core i7-4790 CPU, 3.6 GHz, 4 cores.