predict.cv_sgdnet: Make Predcitions Based on Fit From 'cv_sgdnet()'

In jolars/sgdnet: Fast Sparse Linear Models for Big Data with SAGA

Description

This function is mostly for convenience, since it makes it easy to take an object from a call to cv_sgdnet() and make predictions based on a lambda chosen from cross-validation results.

Usage

## S3 method for class 'cv_sgdnet'
predict(object, newx, s = c("lambda_1se",
  "lambda_min"), ...)

Arguments

object	a fit from cv_sgdnet()
newx	new data to base predictions on
s	'lambda.1se' chooses predictions based on the model fit to the largest lambda with an error at most one standard deviation away from the fit with the least error; predictions are based on the latter fit if 'lambda.min' is chosen
...	arguments passed on to predict.sgdnet()

Value

Predictions for object given data in newx.

Examples

set.seed(1)
train_ind <- sample(150, 100)
fit <- cv_sgdnet(iris[train_ind, 1:4],
                 iris[train_ind, 5],
                 family = "multinomial",
                 nfolds = 5)
predict(fit, iris[-train_ind, 1:4], s = "lambda_min", type = "class")

jolars/sgdnet documentation built on May 22, 2019, 11:52 p.m.