prox.grouplasso: Proximal operator of the group lasso penalty
In jpvert/apg: Optimization with accelerated proximal gradient

Description Usage Arguments Value Examples

Computes the proximal operator of the group lasso penalty:

h(x) = ∑_{g group} w_g ||x[g]||_2 .

Note that the groups should not overlap.

1	prox.grouplasso(x, t, opts = list(groups = as.list(seq(length(xc)))))

x

The input vector

t

The step size

opts

List of parameters, which can include:

groups : a list of groups, each group is just a sequence of indices of the components that form the group (default: all singletons).
groupweigths : a vector of weights for the groups. If a single number, all groups have the same weight (default 1)

The proximal operator of the group lasso, which is a soft-thresholing operator applied to the restriction of the x to each group.

x <- rnorm(5)
# When groups are all the singletons we recover the L1 (lasso) penalty
prox.grouplasso(f,1,list(groups=as.list(seq(length(f)))))
prox.elasticnet(f,1,list(lambda=1,alpha=1))