predict.msgl: Predict

In nielsrhansen/msgl: Multinomial Sparse Group Lasso

Predict

Description

Computes the linear predictors, the estimated probabilities and the estimated classes for a new data set.

Usage

## S3 method for class 'msgl'
predict(object, x, sparse.data = is(x, "sparseMatrix"),
  ...)

Arguments

object	an object of class msgl, produced with msgl.
x	a data matrix of size N_\textrm{new} \times p.
sparse.data	if TRUE x will be treated as sparse, if x is a sparse matrix it will be treated as sparse by default.
...	ignored.

Value

link	the linear predictors – a list of length length(fit$beta) one item for each model, with each item a matrix of size K \times N_\textrm{new} containing the linear predictors.
response	the estimated probabilities – a list of length length(fit$beta) one item for each model, with each item a matrix of size K \times N_\textrm{new} containing the probabilities.
classes	the estimated classes – a matrix of size N_\textrm{new} \times d with d=length(fit$beta).

Author(s)

Martin Vincent

Examples

data(SimData)

x.1 <- x[1:50,]
x.2 <- x[51:100,]

classes.1 <- classes[1:50]
classes.2 <- classes[51:100]

lambda <- msgl::lambda(x.1, classes.1, alpha = .5, d = 50, lambda.min = 0.05)
fit <- msgl::fit(x.1, classes.1, alpha = .5, lambda = lambda)

# Predict classes of new data set x.2
res <- predict(fit, x.2)

# The error rates of the models
Err(res, classes = classes.2)

# The predicted classes for model 20
res$classes[,20]

nielsrhansen/msgl documentation built on Feb. 6, 2024, 1:25 a.m.