Multi-view classification has received considerable attention in recent years. We observed that the existing multi-view classification methods learn a consensus result by collecting all views and thus have two critical limitations. First, it is not scalable. Second, in many applications views of data are available over time; it is infeasible to apply the existing multi-view learning methods to such streaming views. To address the two limitations, in this paper we propose a novel incremental multi-view SVM method, i.e., instead of processing all views simultaneously, we integrate them one by one in an incremental way. We first learn an initial model from the first view; next when a new view is available, we update the model and then apply it to learn a new consensus result. This incremental method is scalable and applicable to streaming views. We present a block coordinate descent algorithm whose convergence is theoretically guaranteed to optimize the induced objective function. Experimental results on several benchmark data sets further demonstrate the effectiveness of our method.
展开▼
