pp_rnv — Robust Normal Variate (RNV)

Group: Preprocessing · Binding: n4m.sklearn.RNV · C ABI: n4m_pp_rnv_*

Description

Robust SNV using median + k * MAD.

Parameters

Name

Type

Default

with_center

bool

True

with_scale

bool

True

k

float

1.4826

Explanations

Bibliographic source

Guo, Q., Wu, W. & Massart, D. L. (1999). The robust normal variate transform for pattern recognition with near-infrared data. Analytica Chimica Acta 382(1–2), 87–103.

Mathematical principle

A median/IQR analogue of SNV: each spectrum is corrected as \((\mathbf{x}_i - \mathrm{median}(\mathbf{x}_i)) / \mathrm{IQR}(\mathbf{x}_i)\). Replacing the mean and standard deviation with robust location/scale estimators makes the normalisation insensitive to a small number of strong absorption bands or outlying channels.

Implementation

C ABI n4m_pp_rnv_* in libn4m (create / apply / destroy lifecycle), wrapped by n4m.sklearn.RNV. The same numerical kernel backs every language binding.

Usage

from n4m.sklearn import RNV
op = RNV()
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).

BackendParity50×250 (ms)250×50 (ms)
C++ native · libn4m
pls4all.cpp.blas✓ ref
pls4all.cpp.blas+omp✓ ref0.20 ms0.22 ms
pls4all.cpp.omp✓ ref
pls4all.cpp.ref✓ ref0.21 ms0.23 ms
Python · pls4all
pls4all.python✓ bind0.19 ms0.20 ms
pls4all.sklearn✓ bind0.19 ms🏆0.19 ms🏆
Python · external
nirs4allsource0.34 ms0.37 ms
ref.python_numpysource

See also: methods index · interactive dashboard