Tensor-based restricted kernel machines for multi-view classification

Multi-view learning deals with data that is described through multiple representations, or views. While various real-world can be represented by three more views, several existing multi-view classification methods only handle two Previously proposed usually solve this issue optimizing pairwise combinations of Although numerically deal the it ignores higher order correlations which examined exploring all views simultaneously. In work new approaches are introduced aim to include statistics when available. The model an extension recently Restricted Kernel Machine classifier and assumes shared hidden features for as well a newly tensor. Experimental results show improvement respect state-of-the art methods, both in terms accuracy runtime.