Skip to contents

Fit a linear model with (sparse) group regularisation

Source: R/group_sparse_lm.R, R/group_sparse_lm_aliases.R
group_sparse_lm.Rd

Adjust a linear model with (sparse) group regularization, that is, a mixture of an element-wise \(\ell_1\) norm and a group-wise mixed-norm (either \(\ell_1/\ell_2\), \(\ell_1/\ell_\infty\) or cooperative). We also add a (possibly structured) \(\ell_2\)-norm (ridge-like). The solution path is computed on an automatically tuned grid of values for the sparse group penalty. The mixture coefficient and the amount of ridge-like regularization are fixed by the user. See details for the criterion optimized.

Usage

group_sparse_lm(
  x,
  y,
  group,
  type = c("l2", "coop", "linf"),
  lambda1 = NULL,
  lambda2 = 0.01,
  alpha = 0,
  weights = rep(1, nrow(x)),
  penscale = sqrt(tabulate(group)),
  struct = Matrix::Diagonal(ncol(x), 1),
  intercept = TRUE,
  normalize = TRUE,
  refit = FALSE,
  nlambda1 = ifelse(is.null(lambda1), 100, length(lambda1)),
  minratio = 0.01,
  maxfeat = ifelse(lambda2 < 0.01, min(2 * nrow(x), ncol(x)), min(4 * nrow(x), ncol(x))),
  beta0 = numeric(ncol(x)),
  control = list()
)

group_lasso(
  x,
  y,
  group,
  lambda1 = NULL,
  lambda2 = 0,
  weights = rep(1, nrow(x)),
  penscale = sqrt(tabulate(group)),
  intercept = TRUE,
  normalize = TRUE,
  nlambda1 = ifelse(is.null(lambda1), 100, length(lambda1)),
  minratio = 0.01,
  maxfeat = ncol(x),
  beta0 = numeric(ncol(x)),
  control = list(method = "quadra")
)

group_l1linf(
  x,
  y,
  group,
  lambda1 = NULL,
  lambda2 = 0,
  weights = rep(1, nrow(x)),
  penscale = sqrt(tabulate(group)),
  intercept = TRUE,
  normalize = TRUE,
  nlambda1 = ifelse(is.null(lambda1), 100, length(lambda1)),
  minratio = 0.01,
  maxfeat = ncol(x),
  beta0 = numeric(ncol(x)),
  control = list()
)

coop_lasso(
  x,
  y,
  group,
  lambda1 = NULL,
  lambda2 = 0,
  weights = rep(1, nrow(x)),
  penscale = sqrt(tabulate(group)),
  intercept = TRUE,
  normalize = TRUE,
  nlambda1 = ifelse(is.null(lambda1), 100, length(lambda1)),
  minratio = 0.01,
  maxfeat = ncol(x),
  beta0 = numeric(ncol(x)),
  control = list()
)

sparse_group_lasso(
  x,
  y,
  group,
  lambda1 = NULL,
  lambda2 = 0,
  alpha = 0.5,
  weights = rep(1, nrow(x)),
  penscale = sqrt(tabulate(group)),
  intercept = TRUE,
  normalize = TRUE,
  nlambda1 = ifelse(is.null(lambda1), 100, length(lambda1)),
  minratio = 0.01,
  maxfeat = ncol(x),
  beta0 = numeric(ncol(x)),
  control = list()
)

sparse_group_l1linf(
  x,
  y,
  group,
  lambda1 = NULL,
  lambda2 = 0,
  alpha = 0.5,
  weights = rep(1, nrow(x)),
  penscale = sqrt(tabulate(group)),
  intercept = TRUE,
  normalize = TRUE,
  nlambda1 = ifelse(is.null(lambda1), 100, length(lambda1)),
  minratio = 0.01,
  maxfeat = ncol(x),
  beta0 = numeric(ncol(x)),
  control = list()
)

sparse_coop_lasso(
  x,
  y,
  group,
  lambda1 = NULL,
  lambda2 = 0,
  alpha = 0.5,
  weights = rep(1, nrow(x)),
  penscale = sqrt(tabulate(group)),
  intercept = TRUE,
  normalize = TRUE,
  nlambda1 = ifelse(is.null(lambda1), 100, length(lambda1)),
  minratio = 0.01,
  maxfeat = ncol(x),
  beta0 = numeric(ncol(x)),
  control = list()
)

Arguments

x

matrix of features, possibly sparsely encoded (experimental). Do NOT include intercept. When normalized is TRUE, coefficients will then be rescaled to the original scale.

y

response vector.

group

vector of integers indicating group belonging. Must match the number of column in x. Must be SORTED integers starting from 1.

type

string indicating the sparse-group variant to be fitted. Could be "l2", "coop", or "linf". Default is "l2" (regular Group-Lasso)

lambda1

sequence of decreasing \(\ell_1\)-penalty levels. If NULL (the default), a vector is generated with nlambda1 entries, starting from a guessed level lambda1.max where only the intercept is included, then shrunken to minratio*lambda1.max.

lambda2

real scalar; tunes the \(\ell_2\) penalty in the Elastic-net. Default is 0.01. Set to 0 to recover the Lasso.

alpha

real scalar in (0,1); tunes mixture between \(\ell_1\) group penalties. Default is 0.0 (standard group-lasso).

weights

vector with real positive values that weight the observations (like in weighted least square). Default sets all weights to 1.

penscale

vector with real positive values that weight the penalty of each feature. Default sets all weights to 1.

struct

matrix structuring the coefficients, possibly sparsely encoded. Must be at least positive semidefinite (this is checked internally). If NULL (the default), the identity matrix is used. See details below.

intercept

logical; indicates if an intercept should be included in the model. Default is TRUE.

normalize

logical; indicates if variables should be normalized to have unit L2 norm before fitting. Default is TRUE.

refit

logical: indicates if the non null coefficients should be refit to avoid excessive bias. Default is FALSE. Can be changed later (both raw and refit coefficients are stored).

nlambda1

integer that indicates the number of values to put in the lambda1 vector. Ignored if lambda1 is provided.

minratio

minimal value of \(\ell_1\)-part of the penalty that will be tried, as a fraction of the maximal lambda1 value. A too small value might lead to instability at the end of the solution path corresponding to small lambda1 combined with \(\lambda_2=0\). The default value tries to avoid this, adapting to the '\(n<p\)' context. Ignored if lambda1 is provided.

maxfeat

integer; limits the number of features ever to enter the model; i.e., non-zero coefficients for the Elastic-net: the algorithm stops if this number is exceeded and lambda1 is cut at the corresponding level. Default is min(nrow(x),ncol(x)) for small lambda2 (<0.01) and min(4*nrow(x),ncol(x)) otherwise. Use with care, as it considerably changes the computation time.

beta0

a starting point for the vector of parameter. By default, will initialized zero. May save time in some situation.

control

list of argument controlling low level options of the algorithm –use with care and at your own risk– :

  • verbose: integer; activate verbose mode –this one is not too risky!– set to 0 for no output; 1 for warnings only, and 2 for tracing the whole progression. Default is 1. Automatically set to 0 when the method is embedded within cross-validation or stability selection.

  • timer: logical; use to record the timing of the algorithm. Default is FALSE.

  • maxiter the maximal number of iteration used in the active set algorithm to solve the problem for a given value of lambda1 . Default is 50.

  • method a string for the underlying solver used. Either "quadra", "fista" or "pgd". Default is "quadra".

  • factmat Boolean indicating if matrix factorization should be used to solve the sub-system. If TRUE (the default), a Cholesky decomposition is maintained along the path. If FALSE, the sub-system are solved with a conjugate gradient algorithm.

  • threshold a threshold for convergence. The algorithm stops when the optimality conditions are fulfill up to this threshold. Default is 1e-6.

  • monitor indicates if a monitoring of the convergence should be recorded, by computing a lower bound between the current solution and the optimum: when '0' (the default), no monitoring is provided; when '1', the bound derived in Grandvalet et al. is computed; when '>1', the Fenchel duality gap is computed along the algorithm.

Value

an object with class SparseGroupFit, inheriting from QuadrupenFit.

Details

The optimized criterion is the following: βhatλ1,λ2 = argminβ 1/2 RSS(β) + λ1 [(1-α) Ωg(β) + α | D β |1] + λ2/2 βT S β, where \(\Omega_g\) is the group-wise mixed norm: \(\ell_1/\ell_2\) (Group-Lasso), \(\ell_1/\ell_\infty\), or cooperative (Clime), controlled by the type argument; \(D\) is a diagonal matrix whose diagonal terms are given by penscale; \(\alpha\) tunes the mixture between the group and element-wise penalties; and \(S\) is the \(\ell_2\) structuring matrix provided via struct, a positive semidefinite matrix (possibly of class Matrix).

See also

See also QuadrupenFit

Examples

## Simulating multivariate Gaussian with blockwise correlation
## and piecewise constant vector of parameters
beta <- rep(c(0,1,0,-1,0), c(25,10,25,10,25))
grp  <- rep(1:5, c(25,10,25,10,25))
cor <- 0.75
Soo <- toeplitz(cor^(0:(25-1))) ## Toeplitz correlation for irrelevant variables
Sww  <- matrix(cor,10,10) ## bloc correlation between active variables
Sigma <- Matrix::bdiag(Soo,Sww,Soo,Sww,Soo)
diag(Sigma) <- 1
n <- 50
x <- as.matrix(matrix(rnorm(95*n),n,95) %*% chol(Sigma))
y <- 10 + x %*% beta + rnorm(n,0,10)

## Various sparse group linear models without/with an additional l2 regularization term
## and with structuring prior
labels <- rep("irrelevant", length(beta))
labels[beta != 0] <- "relevant"

## Group-Lasso
plot(group_lasso(x, y, grp), label=labels)


## Sparse Group-Lasso
plot(sparse_group_lasso(x, y, grp, alpha = 0.75), label=labels)


if (FALSE) { # \dontrun{

## Sparse Group-Lasso + L2 regularization
plot(group_sparse_lm(x, y, grp, type = "l2", alpha = .75, lambda2=.5),
 label=labels)
plot(group_sparse_lm(x, y, grp, type = "l2", alpha = .75, lambda2=10),
 label=labels)
plot(group_sparse_lm(x, y, grp, type = "l2", alpha = .75, lambda2=10,
 struct=solve(Sigma)), label=labels)

## Group-Lasso L1/LINF
plot(group_l1linf(x, y, grp), label=labels)

## Sparse Group-Lasso L1/LINF
plot(sparse_group_l1linf(x, y, grp, alpha = 0.75), label=labels)

## Sparse L1/LINF Group-Lasso + L2 regularization
plot(group_sparse_lm(x, y, grp, type = "linf", alpha = .75, lambda2=.5),
  label=labels)
plot(group_sparse_lm(x, y, grp, type = "linf", alpha = .75, lambda2=10),
  label=labels)
plot(group_sparse_lm(x, y, grp, type = "linf", alpha = .75, lambda2=10,
  struct=solve(Sigma)), label=labels)

## Cooperative-Lasso
plot(coop_lasso(x, y, grp), label=labels)

## Sparse Cooperative-Lasso
plot(sparse_coop_lasso(x, y, grp, alpha = 0.75), label=labels)

## Sparse Cooperative-Lasso + L2 regularization
plot(group_sparse_lm(x, y, grp, type = "coop", alpha = .75, lambda2=.5),
 label=labels)
plot(group_sparse_lm(x, y, grp, type = "coop", alpha = .75, lambda2=10),
 label=labels)
plot(group_sparse_lm(x, y, grp, type = "coop", alpha = .75, lambda2=10,
 struct=solve(Sigma)), label=labels)
} # }