pp_normalize — Column-wise normalization¶
Group: Preprocessing · C ABI: n4m_transform_normalize_*
Description¶
Column-wise normalisation.
Parameters¶
Name |
Type |
Default |
|---|---|---|
|
|
|
|
|
|
API and bindings¶
C ABI (ABI 2): n4m_transform_normalize_create · n4m_transform_normalize_destroy · n4m_transform_normalize_transform. Use the linked public header for the exact signature, configuration, and result handles.
Python (verified public re-export):
from n4m.transform.scaling import Normalize
Source signature: Normalize(feature_min: float = -1.0, feature_max: float = 1.0).
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 canonical paper: the default is column L2 normalization; a non-default feature range selects column-wise min–max scaling. The source defines this mode switch.
Mathematical principle¶
With the default range (-1, 1), every column is divided by \(\sqrt{\sum_i X_{ij}^2}\). If either range endpoint is changed, each column is instead mapped affinely from its observed minimum and maximum to the requested interval.
Appropriate uses¶
Equalizing feature magnitudes before algorithms sensitive to Euclidean scale, or mapping every wavelength to a prescribed numeric range.
Limits and validation¶
The name hides two different operations. Statistics are computed from the matrix passed to the operation rather than stored as a fitted training state, and zero-norm or constant columns can produce non-finite values.
Implementation¶
n4m.transform.scaling.Normalize wraps n4m_transform_normalize_*; the exact default-mode branch is in preprocessing/scaling/normalize.c.
The ABI-2 implementation is the n4m_transform_normalize_* lifecycle in libn4m.
Sources and provenance¶
See also: methods index · interactive dashboard