Multi-View Least Squares Support Vector Machines Classification

In multi-view learning, data is described using different representations, or views. Multi-view classification methods try to exploit information from all views to improve the classification performance. Here a new model is proposed that performs classification when two or more views are available. The model is called Multi-View Least Squares Support Vector Machines (MV-LSSVM) Classification and is based on solving a constrained optimization problem. The primal objective includes a coupling term, which minimizes a combination of the errors from all views. The model combines the benefits from both early and late fusion, it is able to incorporate information from all views in the training phase while still allowing for some degree of freedom to model the views differently. Experimental comparisons with similar methods show that using multiple views improves the results with regard to the single view classifiers and that it outperforms other state-of-the-art algorithms in terms of classification accuracy.