jointseg (version 1.0.2)

multiplyXtXBySparse: multiplyXtXBySparse

Description

Compute C = t(X)*X*val , where val is a row-sparse (n-1)*p matrix and X is the n*(n-1) design matrix for the weighted group fused lasso.

Usage

multiplyXtXBySparse(n, ind, val, w = defaultWeights(n),
  verbose = FALSE)

Arguments

n

Size of the problem

ind

a*1 vector of indices of the non-zero rows of b (each in [1,n-1])

val

a*p matrix whose rows are the non-zero rows of b (same order as ind)

w

(n-1)*1 vector of weights

verbose

A logical value: should extra information be output ? Defaults to FALSE

Value

The (n-1)*p matrix equal to t(X)*X*val

Details

This implementation is derived from the MATLAB code of Vert and Bleakley: http://cbio.ensmp.fr/GFLseg.

References

Bleakley, K., & Vert, J. P. (2011). The group fused lasso for multiple change-point detection. arXiv preprint arXiv:1106.4199. arxiv.org/arXiv:1106.4199

Vert, J. P., & Bleakley, K. (2010). Fast detection of multiple change-points shared by many signals using group LARS. Advances in Neural Information Processing Systems, 23, 2343-2351.

Examples

# NOT RUN {
val <- matrix(c(1.56, 1.35, 1.26, 1.15), ncol=2)
ind <- c(5,6)
n <- 10
res <- multiplyXtXBySparse(n=n, ind=ind, val=val)
res
##           [,1]      [,2]
## [1,] 0.8874235 0.7329904
## [2,] 1.3311352 1.0994855
## [3,] 1.7428651 1.4395645
## [4,] 2.1737347 1.7954524
## [5,] 2.6622704 2.1989711
## [6,] 2.6237347 2.1787857
## [7,] 2.1036678 1.7469149
## [8,] 1.6067028 1.3342283
## [9,] 1.0711352 0.8894855

# }