Ensembles of Partially Trained SVMs with Multiplicative Updates

The training of support vector machines (SVM) involves a quadratic programming problem, which is often optimized by a complicated numerical solver. In this paper, we propose a much simpler approach based on multiplicative updates. This idea was first explored in [Cristianini et al., 1999], but its convergence is sensitive to a learning rate that has to be fixed manually. Moreover, the update rule only works for the hard-margin SVM, which is known to have poor performance on noisy data. In this paper, we show that the multiplicative update of SVM can be formulated as a Bregman projection problem, and the learning rate can then be adapted automatically. Moreover, because of the connection between boosting and Bregman distance, we show that this multiplicative update for SVM can be regarded as boosting the (weighted) Parzen window classifiers. Motivated by the success of boosting, we then consider the use of an adaptive ensemble of the partially trained SVMs. Extensive experiments sho...

• Type: Conference
• Year: 2007
• Where: IJCAI
• Authors: Ivor W. Tsang, James T. Kwok
• Keywords: Artificial Intelligence, IJCAI 2007, Learning Rate, Multiplicative Update, Update Rule