nested.cvSGL: Nested Cross-validation for Sparse Group Lasso

In kouroshz/creNet: Fit a GLM with a combination of lasso and overlapping group lasso regularization

Description

Accurate evaluation of the predictive performance of SGL

Usage

nested.cvSGL(ents, rels, x, y, type = c("linear", "logit"), alphas = seq(0, 1, 0.1), nlam = 20, 
	standardize = c("train", "self", "all", "no"), nfold = 10, measure = c("ll", "auc"), maxit = 1000, 
	thresh = 0.001, min.frac = 0.05, gamma = 0.8, step = 1, reset = 10, cre.sig = 0.01, de.sig = 0.01, 
	ncores = 1, lambdas = NULL, verbose = TRUE)

Arguments

ents	data frame of KB entries
rels	data frame of KB relations
x	matrix of predictors (gene expression levels)
y	response vector (numeric for linear regression, binary for logistic regression)
type	type of model: linear or logistic
alphas	Vector of mixing parameters for lasso/group lasso penalty (0 = pure group lasso, 1 = pure lasso)
nlam	number of lambda values in the regularization path
standardize	type of data standardization to be performed
nfold	number of folds in the inner cross-validation
measure	performance measure used to select best lambdas and alphas: log-likelihood or area under curve
maxit	Maximum number of iterations to convergence
thresh	Convergence threshold for change in beta
min.frac	Minimum value of the penalty parameter, as a fraction of the maximum value
gamma	Fitting parameter used for tuning backtracking (between 0 and 1)
step	Fitting parameter used for initial backtracking step size (between 0 and 1)
reset	Fitting parameter used for taking advantage of local strong convexity in Nesterov momentum (number of iterations before momentum term is reset)
cre.sig
de.sig
ncores	Number of computer cores used in calculations
lambdas	Optional sequence of lambda values for fitting. We recommend leaving this NULL and letting SGL self-select values
verbose	logical for verbosity of display

Value

A list with components

best.lambdas	Best lambda value for each outer fold
best.alphas	Best alpha value for each outer fold
pred	predicted responses (=probabilities for logistic regression)
accuracy	Classification rates
ROC	ROC curve and AUC

Author(s)

David Degras and Kourosh Zarringhalam

kouroshz/creNet documentation built on May 20, 2019, 1:11 p.m.