ga_select — GA-PLS — Genetic Algorithm variable selection

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

Description

GA-PLS genetic algorithm selection

From the pls4all.sklearn.GASelector docstring:

Genetic Algorithm feature selection.

Registry note — R plsVarSel::ga_pls genetic-algorithm variable selection. Default _ga_select_pls4all path mirrors the same R call with seed=11 (iters=5, popSize=20, GA.threshold=3), giving bit-exact mask parity. The C++ splitmix64 GA kernel is opt-in via legacy=True.

Parameters

Name	Type	Default	Notes
n_components	int	2	Number of latent components extracted (k).
n_generations	int	30	Number of GA generations to evolve.
population_size	int	40	Number of candidate feature subsets per generation.
min_features	`int	None`	None
max_features	`int	None`	None
mutation_rate	float	0.05	Per-bit mutation probability applied to GA chromosomes.
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

Leardi, R. (2000). Application of genetic algorithm–PLS for feature selection in spectral data sets. Journal of Chemometrics 14(5–6), 643–655.

Mathematical principle

Wrap a binary genetic algorithm around PLS CV-RMSE. Each chromosome is a \(p\)-bit binary mask encoding which features to include; fitness is \(-\mathrm{CV\text{-}RMSE}\) from PLS on the masked subset; standard GA operators (single-point crossover, bit-flip mutation, elitism) drive the population.

Cost is high — every fitness evaluation is a full PLS fit — but GA-PLS handles non-convex fitness landscapes (non-additive interactions between selected features) that greedy methods miss. Recommended population sizes: 30–100; generations: 100–500.

Stochastic by construction: results vary across RNG seeds. For deterministic comparisons against this selector the benchmark widens the parity tolerance and fixes the seed; in production use a small ensemble of GA runs and take the consensus.

Implementation

n4m_ga_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_ga_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 ga_select_fit
with pls4all.Context() as ctx, pls4all.Config() as cfg:
    res = ga_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 GASelector
mdl = GASelector(n_components=2, n_generations=30, population_size=40, min_features=None, max_features=None, mutation_rate=0.05, n_folds=3, seed=0)
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("ga_select", X, y,
                      n_components = 4L, params = list(n_generations = 5L, population_size = 12L, min_features = 5L, max_features = 20L, mutation_rate = 0.1, seed = 11L))
# res is a named list with MethodResult arrays/scalars.
# selected_indices / top_k_intervals are 1-based.

MATLAB · pls4all (MEX)

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

MATLAB · pls4all (classdef)

No idiomatic classdef wrapper — invoke pls4all.fit("ga_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::ga_pls — genetic-algorithm variable selection. RNG differs from pls4all’s GA so set overlap is loose.

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.

1 thread

Backend	Parity	200×40 (ms)
C++ native · libn4m
pls4all.cpp.blas+omp	✓ J 1.00	607.6 ms
Python · pls4all
pls4all.python	✓ J 1.00	756.5 ms
pls4all.sklearn	⇄ J 0.32	3.94 ms🏆
R · pls4all
pls4all.R	⇄ J 0.32	7.38 ms
pls4all.R.formula	⇄ J 0.32	8.59 ms
pls4all.R.mdatools	⇄ J 0.32	16.0 ms
pls4all.R.pls	⇄ J 0.32	9.76 ms
R · external
📐ref.r_plsvarsel	source	218.7 ms

3 threads

Backend	Parity	200×40 (ms)
C++ native · libn4m
pls4all.cpp.blas+omp	✓ J 1.00	628.1 ms
Python · pls4all
pls4all.python	✓ J 1.00	600.8 ms
pls4all.sklearn	⇄ J 0.32	3.94 ms🏆
R · pls4all
pls4all.R	⇄ J 0.32	7.17 ms
pls4all.R.formula	⇄ J 0.32	7.95 ms
pls4all.R.mdatools	⇄ J 0.32	8.10 ms
pls4all.R.pls	⇄ J 0.32	8.03 ms
R · external
📐ref.r_plsvarsel	source	425.9 ms

10 threads

Backend	Parity	200×40 (ms)
C++ native · libn4m
pls4all.cpp.blas+omp	✓ J 1.00	626.6 ms
Python · pls4all
pls4all.python	✓ J 1.00	1.8 s
pls4all.sklearn	⇄ J 0.32	3.88 ms🏆
R · pls4all
pls4all.R	⇄ J 0.32	7.01 ms
pls4all.R.formula	⇄ J 0.32	7.86 ms
pls4all.R.mdatools	⇄ J 0.32	7.57 ms
pls4all.R.pls	⇄ J 0.32	7.89 ms
R · external
📐ref.r_plsvarsel	source	226.4 ms

---

See also: benchmark overview · methods index · interactive dashboard