Dual Deep Clustering

Deep clustering is a branch of deep learning, in which the dimensionality reduction capabilities of deep networks are exploited for clustering data. In this sense, the deep part of clustering has to be considered more as a preprocessing step than a perfectly integrated module of the neural network. This paper proposes the idea of dual neural network in the framework of gradient-based competitive learning. The theory is based on the intuition that neural networks are able to learn topological structures by working directly on the transpose of the input matrix. In this sense, the dual layer is better suited for handling high-dimensional data, because the weight estimation is driven by a constraining subspace which does not depend on the input dimensionality, but only on the dataset cardinality. This approach allows an exact integration with the deep neural networks in such a way to output the input data prototypes.