wvc_threshold_select — WVC-threshold selection

Group: Variable selector · Registry tolerance: 1e-06

Description

WVC threshold-based selection (§18 Phase 5r)

From the pls4all.sklearn.WVCThresholdSelector docstring:

Threshold-/factor-based WVC-PLS selector.

Registry note — R plsVarSel::WVC_pls with explicit median-scaled threshold. Default _wvc_threshold_select_pls4all path mirrors the same R call, giving bit-exact mask parity. The C++ min-selected kernel is opt-in via legacy=True.

Parameters

Name

Type

Default

Notes

n_components

int

2

Number of latent components extracted (k).

normalize

bool

True

Normalize per-variable scores to sum to one before ranking.

score_threshold

float

0.0

Absolute lower bound on WVC scores for retention.

threshold_factor

float

1.0

Multiplier applied to the mean WVC score to define the dynamic threshold.

min_selected

`int

None`

None

Explanations

Bibliographic source

Andries, J. P. M. & Vander Heyden, Y. (2011). Improved variable reduction in partial least squares modelling based on predictive-property-ranked variables and adaptation of partial least squares complexity. Analytica Chimica Acta 705(1–2), 292–305. https://doi.org/10.1016/j.aca.2011.06.037 — same paper as wvc_select; introduces both the top-\(k\) ranking and the threshold / factor-of-mean rules used here.

Mathematical principle

Apply fixed-threshold and factor-of-mean rules over WVC scores, with a minimum-selected fallback. Two selection rules are evaluated and the lowest-CV-RMSE one returned: (1) WVC > absolute threshold \(\tau\), or (2) WVC > \(f \cdot \overline{\mathrm{WVC}}\) (factor-of-mean).

The factor-of-mean rule is dataset-adaptive; the absolute rule is more conservative. The minimum-selected fallback (e.g. retain at least 10 features) prevents empty selections on flat-WVC datasets.

Implementation

n4m_wvc_threshold_select.

Usage

Every pls4all binding tab dispatches into the same C kernel; the external libraries listed at the bottom of the page are the parity references registered in benchmarks.parity_timing.registry. Switch tabs to read the same fit in your language. The R package now ships drop-in-compatible facades for the CRAN pls package (plsr, pcr, mvr) and for the mdatools::pls(x, y, ...) matrix idiom — those tabs appear only on the methods that have a meaningful equivalence.

pls4all bindings

/* C ABI — libn4m */
n4m_context_t* ctx = n4m_context_create();
n4m_config_t*  cfg = n4m_config_create();
n4m_method_result_t* res = NULL;
n4m_wvc_threshold_select_fit(ctx, cfg, &x_view, &y_view, /* hyperparams */, &res);
/* … read coefficients / mask / scores via */
/* n4m_method_result_get_double_matrix / vector / scalar … */
n4m_method_result_destroy(res);
n4m_config_destroy(cfg);
n4m_context_destroy(ctx);

import pls4all
from pls4all._methods import wvc_threshold_select_fit
with pls4all.Context() as ctx, pls4all.Config() as cfg:
    res = wvc_threshold_select_fit(ctx, cfg, X, y, n_components=4)
# then: res.matrix("predictions"), res.matrix("coefficients"),
# res.vector("mask"), res.scalar("intercept"), …

from pls4all.sklearn import WVCThresholdSelector
mdl = WVCThresholdSelector(n_components=2, normalize=True, score_threshold=0.0, threshold_factor=1.0, min_selected=None)
mdl.fit(X, y)
y_hat = mdl.predict(X_test)

library(pls4all)
# Unified low-level dispatcher (May 2026 R cleanup):
res <- pls4all_method("wvc_threshold_select", X, y,
                      n_components = 4L, params = list(threshold_factor = 0.5, min_selected = 5L))
# res is a named list with MethodResult arrays/scalars.
# selected_indices / top_k_intervals are 1-based.

res  = pls4all.fit("wvc_threshold_select", X, y, "NumComponents", 4);
yhat = predict(res, Xtest);

No idiomatic classdef wrapper — invoke pls4all.fit("wvc_threshold_select", X, y, …) directly from the unified MEX factory.

Registry parity references 📐

  • 📐 ref.r_plsvarsel (R · r) — plsVarSel 0.10.0 · strict (rmse_rel ≤ 1e-06) — R plsVarSel::WVC_pls with explicit threshold — picks features whose weighted-variable scores exceed the median × threshold-factor.

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-06).

Rows tagged with 📐 are the canonical parity references for this method (declared in parity_timing.registry). C++ and external rows show reference parity; pls4all language bindings show binding parity against the C++ backend. Hover the icon for role and tolerance band.

BackendParity200×40 (ms)
C++ native · libn4m
pls4all.cpp.blas+omp✓ J 1.00336.1 ms
Python · pls4all
pls4all.python✓ J 1.00344.7 ms
pls4all.sklearn⇄ J 0.691.67 ms🏆
R · pls4all
pls4all.R⇄ J 0.694.07 ms
pls4all.R.formula⇄ J 0.694.66 ms
pls4all.R.mdatools⇄ J 0.694.59 ms
pls4all.R.pls⇄ J 0.695.51 ms
R · external
📐ref.r_plsvarselsource13.0 ms
BackendParity200×40 (ms)
C++ native · libn4m
pls4all.cpp.blas+omp✓ J 1.00304.2 ms
Python · pls4all
pls4all.python✓ J 1.00301.0 ms
pls4all.sklearn⇄ J 0.691.57 ms🏆
R · pls4all
pls4all.R⇄ J 0.694.31 ms
pls4all.R.formula⇄ J 0.694.42 ms
pls4all.R.mdatools⇄ J 0.694.35 ms
pls4all.R.pls⇄ J 0.694.40 ms
R · external
📐ref.r_plsvarselsource12.7 ms
BackendParity200×40 (ms)
C++ native · libn4m
pls4all.cpp.blas+omp✓ J 1.00340.2 ms
Python · pls4all
pls4all.python✓ J 1.00329.0 ms
pls4all.sklearn⇄ J 0.691.61 ms🏆
R · pls4all
pls4all.R⇄ J 0.694.06 ms
pls4all.R.formula⇄ J 0.694.70 ms
pls4all.R.mdatools⇄ J 0.695.14 ms
pls4all.R.pls⇄ J 0.695.45 ms
R · external
📐ref.r_plsvarselsource13.5 ms

See also: benchmark overview · methods index · interactive dashboard