split_kennard_stone — Kennard Stone Splitter

Group: Splitters · Binding: n4m.sklearn.KennardStoneSplitter · C ABI: n4m_split_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

Explanations

Bibliographic source

Standard spectroscopic operator — see the nirs4all preprocessing / augmentation handbook and the cited literature within the binding docstring.

Mathematical principle

Kennard-Stone train/test split.

Implementation

C ABI n4m_split_kennard_stone_* in libn4m (create / apply / destroy lifecycle), wrapped by n4m.sklearn.KennardStoneSplitter. The same numerical kernel backs every language binding.

Usage

from n4m.sklearn import KennardStoneSplitter
op = KennardStoneSplitter()
X_transformed = op.fit_transform(X)

Benchmarks

Adaptive wall-clock per cell measured against full_matrix.csv. Only backends that implement this method are listed; libraries without the method are omitted.

Verdict  ·  ✓ ref / ≈ ref / ~ shape mark a reference-gate pass at strict / relaxed / qualitative tolerance  ·  ✓ bind = pls4all binding agrees with the C++ baseline  ·  ⇄ cross-check = documented by-design selector/RNG/model, noncanonical API/facade convention, or secondary oracle  ·  ✗ divergent  ·  ⚠ error  ·  — not run. The fastest backend per column is marked 🏆.

Reference gate: strict — numeric equivalence (rmse_rel_tol ≤ 1e-12).

1 thread

Backend	Parity	50×250 (ms)	250×50 (ms)
C++ native · libn4m
pls4all.cpp.blas	✓ ref	0.27 ms	—
pls4all.cpp.blas+omp	✓ ref	0.27 ms	1.10 ms
pls4all.cpp.omp	✓ ref	0.27 ms	—
pls4all.cpp.ref	✓ ref	0.13 ms	0.99 ms
Python · pls4all
pls4all.python	✓ bind	0.11 ms🏆	0.53 ms
pls4all.sklearn	✓ bind	0.11 ms	0.53 ms🏆
Python · external
nirs4all	source	0.39 ms	3.32 ms

---

See also: methods index · interactive dashboard