pp_local_centering — Source-to-target local centering

Group: Signal transforms · C ABI: n4m_transform_local_centering_*

Description

Transfer by subtracting source mean and adding target mean.

Parameters

No constructor parameters.

API and bindings

C ABI (ABI 2): n4m_transform_local_centering_create · n4m_transform_local_centering_destroy · n4m_transform_local_centering_fit · n4m_transform_local_centering_is_fitted · n4m_transform_local_centering_transform. Use the linked public header for the exact signature, configuration, and result handles.

Python (verified public re-export):

from n4m.transform.scatter import LocalCentering

Source signature: LocalCentering().

R: no current source-verified entry point was found for this catalog method.

MATLAB / Octave: no current source-verified entry point was found for this catalog method.

Explanations

Bibliographic source

No unique canonical paper: this is a mean-shift domain-adaptation rule, fully specified by the implementation.

Mathematical principle

Fit computes feature means \(\mu_s\) and \(\mu_t\) from source and target calibration sets. A source row is transferred as \(x-\mu_s+\mu_t\), matching first moments while leaving covariance and higher-order structure unchanged.

Appropriate uses

Correcting simple additive instrument or batch shifts when source and target samples share aligned wavelength columns.

Limits and validation

Only mean shift is corrected. Target means estimated from few or nonrepresentative samples are noisy, and using evaluation target data during fit causes leakage.

Implementation

n4m.transform.scatter.LocalCentering calls n4m_transform_local_centering_*; the paired-domain state and affine shift are in cpp/src/c_api/c_api_advanced.cpp.

The ABI-2 implementation is the n4m_transform_local_centering_* lifecycle in libn4m.

Sources and provenance

https://github.com/GBeurier/nirs4all-methods/blob/main/cpp/src/c_api/c_api_advanced.cpp

---

See also: methods index · interactive dashboard