Learning curve – train / val MSE vs training-set size (R parity)

Manual implementation of the sklearn morie_learning_curve flow: shuffle, split into k folds, for each train-fraction fit on a prefix of the training fold and score on the held-out fold.

Usage

morie_learning_curve(x, y, sizes = NULL, cv = 5L, seed = 0L)

Arguments

x

Numeric matrix predictors.

y

Numeric response.

sizes

Training-set fractions (default seq(0.1, 1.0, length=5)).

cv

Number of CV folds.

seed

RNG seed for shuffling.

Value

Named list: estimate (final val MSE), train_sizes, train_scores, val_scores, n, method.

Examples

set.seed(1)
morie_learning_curve(x = rnorm(50), y = rnorm(50))
#> $estimate
#> [1] 0.9539456
#> 
#> $train_sizes
#> [1]  4 13 22 31 40
#> 
#> $train_scores
#> [1] 0.2567182 0.6734896 0.7965575 0.8384200 0.9145741
#> 
#> $val_scores
#> [1] 1.0013865 0.9445885 0.9452066 0.9623601 0.9539456
#> 
#> $n
#> [1] 50
#> 
#> $method
#> [1] "Learning curve (cv MSE)"
#>