nscancor: Non-Negative and Sparse CCA

Two implementations of canonical correlation analysis (CCA) that are based on iterated regression. By choosing the appropriate regression algorithm for each data domain, it is possible to enforce sparsity, non-negativity or other kinds of constraints on the projection vectors. Multiple canonical variables are computed sequentially using a generalized deflation scheme, where the additional correlation not explained by previous variables is maximized. 'nscancor' is used to analyze paired data from two domains, and has the same interface as the 'cancor' function from the 'stats' package (plus some extra parameters). 'mcancor' is appropriate for analyzing data from three or more domains. See <http://sigg-iten.ch/learningbits/2014/01/20/canonical-correlation-analysis-under-constraints/> and Sigg et al. (2007) <doi:10.1109/MLSP.2007.4414315> for more details.

Getting started

Package details

AuthorChristian Sigg [aut, cre] (<https://orcid.org/0000-0003-1067-9224>), R Core team [ctb] (cancor interface and documentation)
MaintainerChristian Sigg <[email protected]>
LicenseGPL (>= 2)
Version0.6.1-25
URL http://sigg-iten.ch/research/
Package repositoryView on CRAN
Installation Install the latest version of this package by entering the following in R:
install.packages("nscancor")

Try the nscancor package in your browser

Any scripts or data that you put into this service are public.

nscancor documentation built on Feb. 16, 2018, 1:03 a.m.