sipls_select — siPLS — Synergy Interval PLS

Group: Variable selector · Registry tolerance: 0.7 · Parity reference: paper-only (Nørgaard, L., Saudland, A., Wagner, J., Nielsen, J. P., Munck, L. & Engelsen, S. B. (2000). Interval partial least-squares regression (iPLS). Appl. Spectrosc. 54(3), 413-419. (No widely installable R / Python port of siPLS’ synergistic combinations; smoke-tested only.))

Description

siPLS synergistic interval selection (§18 Phase 5q)

From the pls4all.sklearn.SiPLSSelector docstring:

siPLS — synergistic interval combinations (Nørgaard 2000).

Parameters

Name	Type	Default	Notes
n_components	int	2	Number of latent components extracted (k).
interval_width	int	10	Width (in variables) of each contiguous spectral interval.
combination_size	int	2	Number of intervals combined into the synergistic siPLS subset.
n_folds	int	3	Number of cross-validation folds used inside the selector.

Explanations

Bibliographic source

Nørgaard, L., Saudland, A., Wagner, J., Nielsen, J. P., Munck, L. & Engelsen, S. B. (2000). Interval partial least-squares regression (iPLS): a comparative chemometric study with an example from near-infrared spectroscopy. Applied Spectroscopy 54(3), 413–419 — same paper as interval_select; siPLS is the synergy-combinations extension proposed in §3.

Mathematical principle

Exhaustively score every combination of \(m\) fixed-size intervals (out of \(I\) total) by CV-RMSE; return the best combination. This captures synergy — pairs (or triples) of intervals that work well together even if neither alone is the best single interval.

Combinatorial cost is \(\binom{I}{m}\) which is manageable only for small \(m\) (typically \(m \le 4\) with \(I \le 20\), giving up to ~5000 combinations). For larger search spaces use biPLS or interval-GA.

siPLS is the natural extension of iPLS when single-interval performance is unsatisfactory: it explicitly looks for complementary bands.

Implementation

n4m_sipls_select. Reference: R plsVarSel.

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

/* C ABI — libn4m */
n4m_context_t* ctx = n4m_context_create();
n4m_config_t*  cfg = n4m_config_create();
n4m_method_result_t* res = NULL;
n4m_sipls_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);

Python · pls4all (raw)

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

Python · pls4all.sklearn

from pls4all.sklearn import SiPLSSelector
mdl = SiPLSSelector(n_components=2, interval_width=10, combination_size=2, n_folds=3)
mdl.fit(X, y)
y_hat = mdl.predict(X_test)

R · pls4all_method()

library(pls4all)
# Unified low-level dispatcher (May 2026 R cleanup):
res <- pls4all_method("sipls_select", X, y,
                      n_components = 4L, params = list(interval_width = 5L, combination_size = 2L))
# res is a named list with MethodResult arrays/scalars.
# selected_indices / top_k_intervals are 1-based.

MATLAB · pls4all (MEX)

res = pls4all.sipls_select(X, y, 4);
% see header of bindings/matlab/+pls4all/sipls_select.m for full
% parameter surface:
%   res = sipls_select(X, Y, n_components, interval_width, combination_size)
yhat = predict(res, Xtest);

MATLAB · pls4all (classdef)

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

Registry parity references 📐

• 📜 Paper-only — no executable parity reference; the pls4all implementation is verified by a smoke fit only. Canonical citation: Nørgaard, L., Saudland, A., Wagner, J., Nielsen, J. P., Munck, L. & Engelsen, S. B. (2000). Interval partial least-squares regression (iPLS). Appl. Spectrosc. 54(3), 413-419. (No widely installable R / Python port of siPLS’ synergistic combinations; smoke-tested only.)

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 🏆.

1 thread

Backend	Parity	200×40 (ms)
Python · pls4all
pls4all.python	✓ bind	3.18 ms
pls4all.sklearn	✓ bind	2.82 ms🏆
R · pls4all
pls4all.R	✓ bind	7.86 ms
pls4all.R.formula	✓ bind	8.15 ms
pls4all.R.mdatools	✓ bind	8.80 ms
pls4all.R.pls	✓ bind	8.12 ms

3 threads

Backend	Parity	200×40 (ms)
Python · pls4all
pls4all.python	✓ bind	2.97 ms🏆
pls4all.sklearn	✓ bind	3.11 ms
R · pls4all
pls4all.R	✓ bind	6.18 ms
pls4all.R.formula	✓ bind	7.15 ms
pls4all.R.mdatools	✓ bind	7.20 ms
pls4all.R.pls	✓ bind	7.56 ms

10 threads

Backend	Parity	200×40 (ms)
Python · pls4all
pls4all.python	✓ bind	2.57 ms🏆
pls4all.sklearn	✓ bind	2.73 ms
R · pls4all
pls4all.R	✓ bind	6.76 ms
pls4all.R.formula	✓ bind	7.43 ms
pls4all.R.mdatools	✓ bind	18.1 ms
pls4all.R.pls	✓ bind	13.1 ms

---

See also: benchmark overview · methods index · interactive dashboard