ipw_select — IPW — Iterative Predictor Weighting

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

Description

IPW-PLS iterative predictor weighting (§18 Phase 5t)

From the pls4all.sklearn.IPWSelector docstring:

Iterative Predictor Weighting PLS selector.

Registry note — R plsVarSel::ipw_pls iterative predictor weighting (RC filter, scale=TRUE, no.iter=3, IPW.threshold=0.01). Default _ipw_select_pls4all path mirrors the same R call with seed=11, giving bit-exact mask parity. The C++ top-k iterative-score kernel is opt-in via legacy=True.

Parameters

Name

Type

Default

Notes

top_k

int

None

Number of features to retain.

n_components

int

2

Number of latent components extracted (k).

n_iterations

int

20

Number of selection iterations or Monte-Carlo passes.

damping

float

0.5

Exponential moving-average factor mixing previous and current weights in IPW.

weight_floor

float

1e-06

Lower bound applied to per-feature weights to prevent zero-trapping.

n_folds

int

3

Number of cross-validation folds used inside the selector.

seed

int

0

Random seed for reproducible sampling/initialization.

Explanations

Bibliographic source

Forina, M., Casolino, C. & Pizarro Millán, C. (1999). Iterative predictor weighting (IPW) PLS: a technique for the elimination of useless predictors in regression problems. Journal of Chemometrics 13(2), 165–184. https://doi.org/10.1002/(SICI)1099-128X(199903/04)13:2<165::AID-CEM535>3.0.CO;2-Y

Mathematical principle

IPW iteratively re-weights features in \(\mathbf{X}\) by their importance, refits PLS on the re-weighted data, and tracks the score path. Weights are derived from coefficient magnitude after each fit; the iteration converges to a stable importance ranking.

Compared to single-fit coefficient ranking, IPW’s iterative refinement gives more stable rankings when the calibration set is small or noisy. Exposes both the score path (for diagnostic) and the weight path (for interpretation).

Implementation

n4m_ipw_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_ipw_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 ipw_select_fit
with pls4all.Context() as ctx, pls4all.Config() as cfg:
    res = ipw_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 IPWSelector
mdl = IPWSelector(top_k, n_components=2, n_iterations=20, damping=0.5, weight_floor=1e-06, n_folds=3, seed=0)
mdl.fit(X, y)
y_hat = mdl.predict(X_test)

library(pls4all)
# Unified low-level dispatcher (May 2026 R cleanup):
res <- pls4all_method("ipw_select", X, y,
                      n_components = 4L, params = list(n_iterations = 5L, top_k = 4L, damping = 0.5, weight_floor = 0.01))
# res is a named list with MethodResult arrays/scalars.
# selected_indices / top_k_intervals are 1-based.

res = pls4all.ipw_select(X, y, 4);
% see header of bindings/matlab/+pls4all/ipw_select.m for full
% parameter surface:
%   res = ipw_select(X, Y, n_components, n_iterations, top_k, damping, weight_floor)
yhat = predict(res, Xtest);

No idiomatic classdef wrapper — invoke pls4all.fit("ipw_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::ipw_pls — iterative predictor weighting with the RC filter.

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 1.00360.4 ms
Python · pls4all
pls4all.python✓ J 1.00366.9 ms
pls4all.sklearn⇄ J 0.751.81 ms🏆
R · pls4all
pls4all.R⇄ J 0.754.86 ms
pls4all.R.formula⇄ J 0.755.51 ms
pls4all.R.mdatools⇄ J 0.755.15 ms
pls4all.R.pls⇄ J 0.755.86 ms
R · external
📐ref.r_plsvarselsource22.0 ms
BackendParity200×40 (ms)
C++ native · libn4m
pls4all.cpp.blas+omp✓ J 1.00368.1 ms
Python · pls4all
pls4all.python✓ J 1.00358.9 ms
pls4all.sklearn⇄ J 0.751.82 ms🏆
R · pls4all
pls4all.R⇄ J 0.754.32 ms
pls4all.R.formula⇄ J 0.755.47 ms
pls4all.R.mdatools⇄ J 0.755.38 ms
pls4all.R.pls⇄ J 0.755.37 ms
R · external
📐ref.r_plsvarselsource20.7 ms
BackendParity200×40 (ms)
C++ native · libn4m
pls4all.cpp.blas+omp✓ J 1.00392.3 ms
Python · pls4all
pls4all.python✓ J 1.00392.5 ms
pls4all.sklearn⇄ J 0.751.86 ms🏆
R · pls4all
pls4all.R⇄ J 0.754.39 ms
pls4all.R.formula⇄ J 0.755.60 ms
pls4all.R.mdatools⇄ J 0.755.39 ms
pls4all.R.pls⇄ J 0.755.70 ms
R · external
📐ref.r_plsvarselsource26.2 ms

See also: benchmark overview · methods index · interactive dashboard