PUlasso-package {PUlasso} | R Documentation |
PUlasso : An efficient algorithm to solve Positive and Unlabeled(PU) problem with lasso or group lasso penalty
Description
The package efficiently solves PU problem in low or high dimensional setting using Maximization-Minorization and (block) coordinate descent. It allows simultaneous feature selection and parameter estimation for classification. Sparse calculation and parallel computing are supported for the further computational speed-up. See Hyebin Song, Garvesh Raskutti (2018) <https://arxiv.org/abs/1711.08129>.
Details
Main functions: grpPUlasso, cv.grpPUlasso, coef, predict
Author(s)
Hyebin Song, hsong@stat.wisc.edu, Garvesh Raskutti, raskutti@stat.wisc.edu.
See Also
Useful links:
Examples
data("simulPU")
fit<-grpPUlasso(X=simulPU$X,z=simulPU$z,py1=simulPU$truePY1)
## Not run:
cvfit<-cv.grpPUlasso(X=simulPU$X,z=simulPU$z,py1=simulPU$truePY1)
## End(Not run)
coef(fit,lambda=fit$lambda[10])
predict(fit,newdata = head(simulPU$X), lambda= fit$lambda[10],type = "response")
[Package PUlasso version 3.2.5 Index]