ridge_active_superblock — n4m.compose.aom_superblock.ridge_active_superblock

Namespace: n4m.compose.aom_superblock · Fully-qualified: n4m.compose.aom_superblock.ridge_active_superblock · Catalog id: aom_pop.ridge_active_superblock

API surface

C ABI: no standalone exported symbol is declared for this method.

Python (verified public re-export): from n4m.compose.aom_superblock import aom_ridge_active_superblock

Signature: aom_ridge_active_superblock(X, y, *, operators = None, alpha: float | None = None, alphas: Sequence[float] = (0.0001, 0.01, 1.0, 100.0), cv: int = 5, fold_ids = None, active_top_m: int = 20, active_diversity_threshold: float = 0.98, active_score_method: str = 'norm', active_max_per_family: int | None = None, keep_identity: bool = True, block_scaling: str = 'rms', center_x: bool = True, center_y: bool = True)

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.

Parameters

Name

Type

Default

X

—

required

y

—

required

operators

—

None

alpha

float | None

None

alphas

Sequence[float]

(0.0001, 0.01, 1.0, 100.0)

cv

int

5

fold_ids

—

None

active_top_m

int

20

active_diversity_threshold

float

0.98

active_score_method

str

'norm'

active_max_per_family

int | None

None

keep_identity

bool

True

block_scaling

str

'rms'

center_x

bool

True

center_y

bool

True

Explanations

Bibliographic source

No canonical paper defines this active-superblock variant. Beurier, G. et al. (2026). AOM-PLS / POP-PLS paper companion, arXiv:2605.13587, https://arxiv.org/abs/2605.13587. The product variants below are implementation-specific extensions; the paper does not by itself specify their ABI-2 orchestration.

Mathematical principle

Form centered (optionally RMS-scaled) blocks \(Z_b\). With centered response \(Y_c\), each candidate has signature \(s_b=c_bZ_b^\top Y_c\) and default activity score \(\lVert s_b\rVert_F^2\) (KTA and blend scores are optional). The selector keeps top-\(M\) candidates while pruning highly correlated signatures, enforcing family caps/identity, then fits Ridge on the selected concatenated superblock.

Appropriate uses

Sparse or constrained superblock workflows that need a linear Ridge prediction surface and branch diagnostics.

Limits and validation

Activity, signature-correlation pruning, family caps, and Ridge penalty are refit inside each fold. This is not a substitute for external validation and cannot fold non-linear or stateful branches into raw-space coefficients.

Implementation

n4m.compose.aom_superblock.aom_ridge_active_superblock and AOMRidgeActiveSuperblock; no standalone C ABI symbol.

Sources and provenance

https://github.com/GBeurier/nirs4all-methods/blob/main/bindings/python/src/n4m/_impl/native.py; https://github.com/GBeurier/nirs4all-methods/blob/main/bindings/python/src/n4m/compose/aom_superblock.py

Catalog note

Native AOM Ridge active superblock over strict-linear single-operator AOM views (core fit_aom_superblock). Inside every alpha-CV fold it screens operators by response signatures of the actual operator outputs (norm, KTA or blend score, diversity threshold, optional per-family cap, identity kept first), fits Ridge on the active superblock, refits the screen on all rows, and folds the final coefficients back to original-input input_coefficients plus intercept (affine N4MM state). It excludes donor branch_global, MKL/kernel, row-reference-dependent preprocessing and nonlinear AOM Ridge modes.

Timing benchmark: benchmarks/cross_binding/bench_aom_ridge_active_superblock_timing.py

See also: methods index.