Canonical Correlation Analysis With L2, 1-Norm for Multiview Data Representation

Meixiang Xu, Zhenfeng Zhu, Xingxing Zhang, Yao Zhao, Xuelong Li

Abstract

For many machine learning algorithms, their success heavily depends on data representation. In this paper, we present an ℓ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2,1</sub> -norm constrained canonical correlation analysis (CCA) model, that is, L <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2,1</sub> -CCA, toward discovering compact and discriminative representation for the data associated with multiple views. To well exploit the complementary and coherent information across multiple views, the ℓ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2,1</sub> -norm is employed to constrain the canonical loadings and measure the canonical correlation loss term simultaneously. It enables, on the one hand, the canonical loadings to be with the capacity of variable selection for facilitating the interpretability of the learned canonical variables, and on the other hand, the learned canonical common representation keeps highly consistent with the most canonical variables from each view of the data. Meanwhile, the proposed L <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2,1</sub> -CCA can also be provided with the desired insensitivity to noise (outliers) to some degree. To solve the optimization problem, we develop an efficient alternating optimization algorithm and give its convergence analysis both theoretically and experimentally. Considerable experiment results on several realworld datasets have demonstrated that L <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2,1</sub> -CCA can achieve competitive or better performance in comparison with some representative approaches for multiview representation learning.
BibTeX
@article{xu2020canonical,
  author = {Meixiang Xu and Zhenfeng Zhu and Xingxing Zhang and Yao Zhao and Xuelong Li},
  title = {{Canonical Correlation Analysis With L2, 1-Norm for Multiview Data Representation}},
  journal = {IEEE Trans. Cybernetics},
  year = {2020},
  doi = {10.1109/tcyb.2019.2904753},
}

Loading related content...

Figure