This repository is a fork of the original fwelnet
implementation by
Tay et al.
The goal of this fork is support a research project based on the
extension of fwelnet
to survival settings and competing event
analysis. For this purpose, some of the internal functions were modified
in a way that is not directly compatible with the original
implementation, hence this project currently stands on its own, rather
than being a pull-request to the original implementation.
This version, renamed to cooper
, supports targets of the form
Surv(time, event)
from the survival
package, and fits fwelnet
through the minimization of the Cox deviance. In addition, it adds the
cooper()
function for a specific implementation of the multi-task
algorithm proposed by Tay et al. which fits cause-specific Cox models.
As of now, the implementation is tailored towards a specific research
question rather than general use.
Content from the original repository follows.
fwelnet
is a package that fits the feature-weighted elastic net
(fwelnet), a variant of the elastic net which has feature-specific
penalties. These penalties are based on additional information that the
user has on the features. fwelnet
works with continuous and binary
responses. For details, please see the
preprint. For a short tutorial on
how to use the package, please see the vignette in the vignettes/
folder.
Here is a simple example to illustrate how to use this package. First, let’s generate some data. In this example, we assume that we have 40 features, and that these features come in 4 groups of 10. The response is a linear combination of the features from the first 2 groups with additive Gaussian noise.
set.seed(1)
n <- 100
p <- 40
groups <- list(1:10, 11:20, 21:30, 31:40) # which features belong to which group
x <- matrix(rnorm(n * p), nrow = n, ncol = p)
beta <- matrix(rep(1:0, each = 20), ncol = 1)
y <- x %*% beta + rnorm(n)
In order to fit a fwelnet model, we have to define a feature information
matrix. In our example, we have
# generate Z matrix
z <- matrix(0, nrow = p, ncol = length(groups))
for (i in 1:length(groups)) {
z[groups[[i]], i] <- 1
}
Once z
is specified, we can fit the fwelnet model with fwelnet()
.
library(fwelnet)
fit <- fwelnet(x, y, z)
“fwelnet” objects are equipped with predict
and coef
methods which
allow the user to make predictions on new data and to view the model
coefficients. By default predictions and coefficients are returned for
the whole lambda
path.
# predictions for first 5 observations at 20th lambda value
predict(fit, x[1:5, ])[, 20]
# [1] 1.2359876 6.3969524 -1.9602492 -1.1781763 -0.5872078
# coefficients at the 20th lambda value (including intercept)
as.numeric(coef(fit)[, 20])
# [1] -0.1440444 0.6481368 0.5835123 0.5545719 0.7626670 0.9323669
# [7] 0.7142652 0.7729034 1.1863878 0.9510465 0.7041017 0.7242916
# [13] 0.7258272 0.9641891 0.7273089 0.7733351 0.8912650 1.2407252
# [19] 0.5566069 1.1192067 0.7399865 0.0000000 0.0000000 0.0000000
# [25] 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000
# [31] 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000
# [37] 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000