Enum HoldoutStrategy 

pub enum HoldoutStrategy {
    None,
    Periodic {
        period: usize,
    },
    Random {
        holdout_fraction: f64,
        seed: u64,
    },
}

Strategy for deciding which samples are held out (test-only, never trained on).

All strategies evaluate every sample; the distinction is whether the model is trained on it afterward.

Variants

None

Every sample is used for both test and train (equivalent to prequential).

Periodic

Every N-th sample is test-only (not trained on).

The remaining (N-1)/N fraction of samples are used for training.

Fields

period: usize

Holdout period. Every period-th sample is test-only.

Random

Each sample has probability holdout_fraction of being test-only.

Uses a deterministic xorshift64 PRNG seeded with seed for reproducibility. The same seed always produces the same holdout pattern.

Fields

holdout_fraction: f64

Probability that a sample is held out (not trained on).

seed: u64

Seed for the deterministic PRNG.

Trait Implementations

impl Clone for HoldoutStrategy

fn clone(&self) -> HoldoutStrategy

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for HoldoutStrategy

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.