ggcsvd: Groups sparse Generalized SVD of a matrix
In vguillemot/sparseMCA:

Description Usage Arguments Value Author(s) Examples

Groups sparse Generalized SVD of a matrix

ggcsvd(X, R = 2, au = rep(sqrt(nrow(X)), R), av = rep(sqrt(ncol(X)),
  R), Gu = seq(NROW(X)), Gv = seq(NCOL(X)), itermax.pi = 1000,
  itermax.pocs = 1000, eps.pi = 1e-16, eps.pocs = 1e-16,
  init = "svd", order_sv = TRUE)

`X`	a (data) matrix;
`R`	the desired rank of the singular decomposition;
`au`	The radiuses (>0) of the $L_1$ ball for each left vector. Default to the maximum possible radius, such that the result is the same as the result of a regular SVD.
`av`	The radiuses (>0) of the $L_1$ balls for each right vector. Default to the maximum possible radius, such that the result is the same as the result of a regular SVD.
`Gu`	a vector describing the groups for the lines.
`Gv`	a vector describing the groups for the columns.
`itermax.pi`	The maximum number of iterations for the power iteration.
`itermax.pocs`	The maximum number of iterations for POCS.
`eps.pi`	Precision for the power iteration.
`eps.pocs`	Precision for POCS.
`init`	How to initialize the algorithm. Either "svd" (default) or "rand" to intialize with, respectively, the results of a regular SVD or random vectors.
`order_sv`	Boolean. Should the singular values be artificially ordered. Default to TRUE.

Pseudo-singular vectors (U and V) and values (D), and the number of iterations.

Vincent Guillemot

1
2
3

X <- matrix(rnorm(20), 5, 4)
ggcsvd(X) # By default, no sparsity constraint is imposed on the decomposition
svd(X) # Should give roughly the same results