pp_wavelet_pca — PCA of multilevel wavelet coefficients¶
Group: Augmentation · C ABI: n4m_transform_wavelet_pca_*
Description¶
DWT coefficient projection through PCA scores.
Parameters¶
Name |
Type |
Default |
|---|---|---|
|
|
|
|
|
|
|
|
|
|
|
|
API and bindings¶
C ABI (ABI 2): n4m_transform_wavelet_pca_create · n4m_transform_wavelet_pca_destroy · n4m_transform_wavelet_pca_fit · n4m_transform_wavelet_pca_is_fitted · n4m_transform_wavelet_pca_output_cols · n4m_transform_wavelet_pca_transform. Use the linked public header for the exact signature, configuration, and result handles.
Python (verified public re-export):
from n4m.transform.wavelet import WaveletPCA
Source signature: WaveletPCA(family: str = 'haar', mode: str = 'periodization', max_level: int = 2, n_components: float = 5.0).
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 unique paper defines this composite. It combines Mallat’s DWT (https://doi.org/10.1109/34.192463) with Pearson PCA (https://doi.org/10.1080/14786440109462720).
Mathematical principle¶
Each row’s multilevel DWT coefficients are packed into one vector. Fit centers that coefficient matrix, performs compact SVD, and retains an integer component count or the smallest count reaching the requested explained-variance fraction; transform projects centered coefficients onto those loadings.
Appropriate uses¶
Joint multiscale denoising and dimensionality reduction when raw wavelet coefficients are too numerous for the downstream model.
Limits and validation¶
Both DWT and PCA are scale/shift sensitive; interpretation mixes bands and fitting is training-dependent. A variance threshold can yield changing output dimension across folds.
Implementation¶
n4m.transform.wavelet.WaveletPCA wraps n4m_transform_wavelet_pca_*; packing, centering, SVD, selection, and projection are in wavelet_pca.c.
The ABI-2 implementation is the n4m_transform_wavelet_pca_* lifecycle in libn4m.
Sources and provenance¶
https://doi.org/10.1109/34.192463; https://doi.org/10.1080/14786440109462720; https://github.com/GBeurier/nirs4all-methods/blob/main/cpp/src/core/preprocessing/wavelets/wavelet_pca.c
See also: methods index · interactive dashboard