Deep multi-view clustering: A comprehensive survey of the contemporary techniques

Abstract

Data can be represented by multiple sets of features, where each semantically coherent set of features is called a view. For example, an image can be represented by multiple sets of features that measure textures, shapes, edge features, etc. Collecting multiple views of data is generally easier than annotating it with the help of experts. Thus, the unsupervised exploration of data in consultation with all collected views is essential to identify naturally occurring clusters of data instances. In deep multi-view clustering, deep neural networks are used to obtain non-linear latent representations of data instances that agree with the multiple views, using which clusters of data instances are identified. A wide variety of such deep multi-view clustering approaches exist, which we systematically study and categorize into a novel taxonomy that provides structure to the existing literature and can also guide future researchers. We provide a pedagogical discussion on preliminary concepts to help understand topics relevant to the studied deep clustering methods. Various multi-view problems that are being studied are summarized, and future research scopes have been noted.