Unsupervised Multi-View Feature Selection via Co-Regularization

Existing unsupervised feature selection algorithms are designed to extract the most relevant subset of features that can facilitate clustering and interpretation of the obtained results. However, these techniques are not applicable in many real-world scenarios where one has an access to datasets consisting of multiple views/representations e.g. various omics profiles of the patients, medical text records coupled with FMRI images, etc. In this paper, we introduce a novel unsupervised multi-view feature selection algorithm that can simultaneously extract a relevant subset of features and perform clustering that is consistent across different views. By leveraging information from these different views we obtain more robust and accurate results in comparison to traditional methods. Our algorithm works by co-regularizing the multiple clustering hypothesis and at the same time learning the feature ranking matrix. In the empirical evaluation on high dimensional multi-view metagenome data we demonstrate the applicability and efficacy of the proposed algorithm. Last but not least, a Python implementation of our algorithm is available for download, easy to setup and use.1