D-GCCA: Decomposition-based Generalized Canonical Correlation Analysis for Multi-view High-dimensional Data
This python package implements the D-GCCA method proposed in [1]. See example.py for details, with Python 3.
D-GCCA decomposes the observed multi-view data into
where are low-rank common-source matrices that represent the signal data coming from the common latent factors shared across all data views, and
are low-rank distinctive-source matrices each from the distinctive latent factors of the corresponding view, and
are noise matrices.
In other words, the common-source and distinctive-source matrices contain the variation information in each view, respectively, explained by the common and distinctive latent factors of the K views.
Please cite the article [1] for this package.
For K=2 data views, please use the D-CCA package, instead.
[1] Shu, H., Qu, Z., & Zhu, H. "D-GCCA: Decomposition-based Generalized Canonical Correlation Analysis for Multi-view High-dimensional Data". Journal of Machine Learning Research, 23(169):1−64