Learning curves: train/validation RMSE vs training-set size (Geron Ch 4, morie.fn hmlcv)

Split order via the shared LCG stream, not numpy PCG64 permutation.

Usage

morie_geron_learning_curves_hmlcv(
  X,
  y,
  n_splits = 10,
  val_fraction = 0.2,
  fit = NULL,
  predict = NULL,
  seed = 0
)

Arguments

X, y

Data.

n_splits

Sizes to evaluate.

val_fraction

Held-out fraction.

fit, predict

Optional callables (default OLS via lstsq).

seed

Split seed.

Value

List with train_sizes, rmse_train, rmse_val, final_gap, verdict, estimate, n, method.

Examples

V <- c(1, 2, 3, 4, 5, 6, 7, 8)
morie_geron_learning_curves_hmlcv(V, V)
#> $train_sizes
#> [1] 2 3 4 5 6
#> 
#> $rmse_train
#> [1] 0.000000e+00 1.538370e-15 1.601186e-15 8.188600e-16 7.691851e-16
#> 
#> $rmse_val
#> [1] 0.000000e+00 3.510833e-16 3.510833e-16 1.755417e-16 1.755417e-16
#> 
#> $final_gap
#> [1] -5.936434e-16
#> 
#> $verdict
#> [1] "fits well"
#> 
#> $estimate
#> [1] 1.755417e-16
#> 
#> $n
#> [1] 8
#> 
#> $method
#> [1] "Learning curves (train/validation RMSE vs training-set size)"
#>