The task of identifying important biomarkers from multiple groups of medical examinations is formulated as a multi-view feature selection problem. We utilize the tensor product operation to model feature interactions across different data sources, factorization techniques to reduce the optimization, and recursive feature elimination to select discriminative biomarkers. The proposed method can work efficiently with many views and effectively with both linear and nonlinear kernels.
© Bokai Cao, 2018. Licensed under an Apache-2 license.
Bokai Cao, Lifang He, Xiangnan Kong, Philip S. Yu, Zhifeng Hao and Ann B. Ragin. Tensor-based Multi-view Feature Selection with Applications to Brain Diseases. In ICDM 2014.