emil (version 2.0.2)

learning_curve: Learning curve analysis

Description

This function studies the change in permformance as the sizes of the fitting and test sets are varied. In case the studied modeling procedures cannot produce models on the smallest fitting sets, please use .return_error=TRUE (see evaluate.

Usage

learning_curve(procedure, x, y, fraction, nfold = 100, ..., .verbose = TRUE)

Arguments

procedure

modeling_procedure.

x

Dataset descriptors.

y

Response.

fraction

Fraction of dataset to hold out, i.e. use as test set. Defaults 20 logarithmically distributed values ranging from all but 5 observations per class in the largest test set to only 5 observations per class in the smallest test set.

nfold

How many holdout folds that should be calculated.

...

Sent to evaluate.

.verbose

Whether to print an activity log. Set to -1 to also suppress output generated from the procedure's functions.

References

Richard O Duda, Peter E Hart, and David G Stork. Pattern Classification. Wiley, 2nd edition, 2000. ISBN 978-0-471-05669-0.

Examples

procedure <- lapply(c(Linear="lda", Quadratic="qda"), modeling_procedure)
options(emil_max_indent=3)
lc <- learning_curve(procedure, iris[-5], iris$Species, fraction=7:1/10)
plot(lc)