`split_kennard_stone` — Kennard–Stone maximin calibration split¶

Group: Splitters · C ABI: n4m_model_selection_kennard_stone_*

Description¶

Kennard-Stone train/test split.

Full binding docstring

Kennard-Stone train/test split.

Picks the most diverse samples for the training set, in descending order
of pairwise Euclidean distance.

Parameters¶

Name	Type	Default
`test_size`	`float`	`0.25`

API and bindings¶

C ABI (ABI 2): n4m_model_selection_kennard_stone_create · n4m_model_selection_kennard_stone_destroy · n4m_model_selection_kennard_stone_split. Use the linked public header for the exact signature, configuration, and result handles.

Python (verified public re-export):

from n4m.model_selection.splitters import KennardStone

Source signature: KennardStone(test_size: float = 0.25).

R (source-verified): kennard_stone_split(X, test_size = 0.25, zero_based = FALSE).

library(n4m)
result <- kennard_stone_split(X)

MATLAB / Octave (source-verified): kennard_stone_split(X, varargin).

addpath('bindings/matlab')
result = n4m.kennard_stone_split(X);

Explanations¶

Bibliographic source¶

Kennard & Stone (1969), Computer Aided Design of Experiments, Technometrics 11, 137–148, DOI 10.1080/00401706.1969.10490666 (https://doi.org/10.1080/00401706.1969.10490666).

Mathematical principle¶

Compute all Euclidean distances in X, initialize the training set with the globally farthest pair, then repeatedly add the unselected sample maximizing its minimum distance to the selected set. For n samples, test_size gives \(n_{test}=\lceil n\,test\_size\rceil\) and the remaining maximin points are training.

Appropriate uses¶

Constructing a calibration set that covers X-space while reserving the complement for testing.

Limits and validation¶

It is deterministic, quadratic in memory/time for distances, sensitive to feature scaling, and ignores y. The test complement is not an IID random sample.

Implementation¶

Python role API n4m.model_selection.splitters.KennardStone; binding class KennardStoneSplitter; ABI 2 family n4m_model_selection_kennard_stone_{create,split,destroy}.

The ABI-2 implementation is the n4m_model_selection_kennard_stone_* lifecycle in libn4m.

Sources and provenance¶

https://github.com/GBeurier/nirs4all-methods/blob/main/cpp/src/core/splitters/kennard_stone.h; https://doi.org/10.1080/00401706.1969.10490666

See also: methods index · interactive dashboard

split_kennard_stone — Kennard–Stone maximin calibration split¶