Repository Public
1// =============================================================================2// SP BOUCHER | SOFTWARE & RESEARCH3// ARTIFICIAL INTELLIGENCE • FINANCE • ENGINEERING4// =============================================================================5//6// PROJECT METADATA7// -----------------------------------------------------------------------------8// Project : ORIX Sheet9// File : Packages/SheetAnalytics/Tests/StatisticsTests/LeastSquaresTests.swift10// Module : StatisticsTests11// Repository : https://git.spboucher.ai/orix-sheet12// Created : 2026-11-2113// Last Modified : 2026-11-2114//15// =============================================================================16// AUTHOR & PROFESSIONAL AFFILIATION17// =============================================================================18//19// Author : Simon-Pierre Boucher20// Position : Professor of Finance21// Institution : Université du Québec en Outaouais (UQO)22// Department : Département des sciences administratives23// Organization : SP Boucher24//25// Website : https://spboucher.ai26// Git : https://git.spboucher.ai27//28// =============================================================================29// OFFICIAL CONTACT DIRECTORY30// =============================================================================31//32// General : info@spboucher.ai33// Professional : contact@spboucher.ai34// Direct : simon@spboucher.ai35// Research : research@spboucher.ai36// Development : dev@spboucher.ai37// Support : support@spboucher.ai38// Security/Admin : admin@spboucher.ai39// Notifications : notifications@spboucher.ai40//41// Preferred Contact: support@spboucher.ai42//43// =============================================================================44// COPYRIGHT & INTELLECTUAL PROPERTY45// =============================================================================46//47// Copyright (c) 202648// Simon-Pierre Boucher49//50// All rights reserved where applicable.51//52// This file is governed by the applicable project license,53// copyright notices, institutional agreements and relevant law.54//55// Unless expressly authorized by the applicable license or56// rights holder, copying, redistribution, sublicensing,57// publication and commercial exploitation are not permitted.58//59// Existing third-party rights and notices remain unaffected.60//61// =============================================================================62// CONFIDENTIALITY & SECURITY63// =============================================================================64//65// Classification : Public66// Distribution : Public repository; use is governed by the project license67// Security Contact : admin@spboucher.ai68//69// Access to this source does not independently grant permission70// to redistribute or disclose confidential materials.71//72// Do not embed credentials, private keys, access tokens,73// certificates or confidential personal information.74//75// =============================================================================76// LICENSE & MAINTENANCE77// =============================================================================78//79// License : None specified (no license file in this repository)80// Maintainer : Simon-Pierre Boucher81// Technical Contact: dev@spboucher.ai82//83// =============================================================================84// END OF SP BOUCHER SOURCE HEADER85// =============================================================================8687import Foundation88import Testing89import AnalyticsTestSupport90@testable import Statistics9192private let lsFixture = Fixture.load("multivariate", .module)["leastsq"]9394@Suite struct LeastSquaresTests {95 func checkCovariance(_ r: LeastSquaresResult, _ ref: JSON, rel: Double, _ label: String) {96 let want = ref["pcov"].mat97 guard let cov = r.covariance else { Issue.record("\(label): no covariance"); return }98 #expect(allClose(cov.data, want.flatMap { $0 }, rel: rel), "\(label) covariance \(maxRelativeError(cov.data, want.flatMap { $0 }))")99 #expect(allClose(r.standardErrors!, want.indices.map { want[$0][$0].squareRoot() }, rel: rel), "\(label) se")100 }101102 @Test func nistMisra1a() throws {103 let f = lsFixture["misra1a"]104 let model: (Double, [Double]) -> Double = { x, b in b[0] * (1 - exp(-b[1] * x)) }105 for startKey in ["start1", "start2"] {106 let r = try LeastSquares.curveFit(model: model, x: f["x"].arr, y: f["y"].arr, start: f[startKey].arr)107 #expect(r.converged, "\(startKey): \(r.termination)")108 #expect(allClose(r.parameters, f["certified"]["b"].arr, rel: 1e-8), "\(startKey) vs NIST: \(r.parameters)")109 #expect(isClose(r.sumOfSquares, f["certified"]["rss"].d, rel: 1e-9))110 #expect(allClose(r.parameters, f["popt"].arr, rel: 1e-8))111 checkCovariance(r, f, rel: 1e-6, startKey)112 #expect(r.degreesOfFreedom == 12)113 #expect(isClose(r.residualVariance, r.sumOfSquares / 12, rel: 1e-14))114 }115 }116117 @Test func nistDanWood() throws {118 let f = lsFixture["danwood"]119 let model: (Double, [Double]) -> Double = { x, b in b[0] * pow(x, b[1]) }120 for startKey in ["start1", "start2"] {121 let r = try LeastSquares.curveFit(model: model, x: f["x"].arr, y: f["y"].arr, start: f[startKey].arr)122 #expect(r.converged)123 #expect(allClose(r.parameters, f["certified"]["b"].arr, rel: 1e-8), "\(startKey) vs NIST: \(r.parameters)")124 #expect(isClose(r.sumOfSquares, f["certified"]["rss"].d, rel: 1e-9))125 checkCovariance(r, f, rel: 1e-6, startKey)126 }127 }128129 @Test func exponentialDecayAndJacobianVariants() throws {130 let f = lsFixture["decay"]131 let x = f["x"].arr, y = f["y"].arr132 let resid: ([Double]) -> [Double] = { p in zip(x, y).map { p[0] * exp(-p[1] * $0) + p[2] - $1 } }133 let analytic: ([Double]) -> Matrix = { p in134 Matrix(x.map { [exp(-p[1] * $0), -p[0] * $0 * exp(-p[1] * $0), 1] })135 }136 var forward = LeastSquares.Options()137 forward.jacobian = .forward138 let a = try LeastSquares.levenbergMarquardt(residuals: resid, start: f["start"].arr, jacobian: analytic)139 let c = try LeastSquares.levenbergMarquardt(residuals: resid, start: f["start"].arr)140 let fw = try LeastSquares.levenbergMarquardt(residuals: resid, start: f["start"].arr, options: forward)141 for (r, label, tol) in [(a, "analytic", 1e-8), (c, "central", 1e-8), (fw, "forward", 1e-6)] {142 #expect(r.converged, "\(label)")143 #expect(allClose(r.parameters, f["popt"].arr, rel: tol), "\(label) \(r.parameters)")144 #expect(isClose(r.sumOfSquares, f["rss"].d, rel: 1e-9), "\(label) rss")145 }146 checkCovariance(a, f, rel: 1e-7, "analytic")147 checkCovariance(c, f, rel: 1e-6, "central")148 checkCovariance(fw, f, rel: 1e-4, "forward")149 #expect(a.functionEvaluations < c.functionEvaluations)150 #expect(a.gradientNorm < 1e-6)151 #expect(a.table().count == 10)152 }153154 @Test func weightedGaussianPeak() throws {155 let f = lsFixture["gauss"]156 let sigma = f["sigma"].arr157 let r = try LeastSquares.curveFit(model: { x, p in p[0] * exp(-0.5 * pow((x - p[1]) / p[2], 2)) },158 x: f["x"].arr, y: f["y"].arr, start: f["start"].arr, weights: sigma.map { 1 / $0 })159 #expect(r.converged)160 #expect(allClose(r.parameters, f["popt"].arr, rel: 1e-8))161 checkCovariance(r, f, rel: 1e-6, "gauss")162 }163164 @Test func rosenbrockAndLinearProblems() throws {165 let r = try LeastSquares.levenbergMarquardt(residuals: { p in [10 * (p[1] - p[0] * p[0]), 1 - p[0]] }, start: [-1.2, 1])166 #expect(r.converged)167 #expect(allClose(r.parameters, [1, 1], rel: 1e-9)) // analytic minimum168 #expect(allClose(r.parameters, lsFixture["rosenbrock"]["x"].arr, rel: 1e-8))169 #expect(r.sumOfSquares < 1e-18)170 #expect(r.covariance == nil && r.degreesOfFreedom == 0) // n = p: no residual variance171 // a linear model is solved exactly: y = 2 + 3x with one outlier-free fit172 let xs = [0.0, 1, 2, 3, 4], ys = [2.1, 4.9, 8.2, 10.9, 14.1]173 let lin = try LeastSquares.curveFit(model: { x, p in p[0] + p[1] * x }, x: xs, y: ys, start: [0, 0])174 let slope = Descriptive.covariance(xs, ys) / Descriptive.variance(xs)175 #expect(allClose(lin.parameters, [Descriptive.mean(ys) - slope * 2, slope], rel: 1e-10))176 // OLS standard error of the slope: s / √Sxx177 #expect(isClose(lin.standardErrors![1], (lin.residualVariance / Descriptive.sumSquaredDeviations(xs)).squareRoot(), rel: 1e-8))178 // already at the optimum179 let done = try LeastSquares.levenbergMarquardt(residuals: { p in [p[0] - 1, p[1] + 2] }, start: [1, -2])180 #expect(done.converged && done.iterations == 0 && done.termination == "gradient")181 }182183 @Test func invalidInputs() {184 #expect(throws: StatError.self) { try LeastSquares.levenbergMarquardt(residuals: { _ in [1] }, start: []) }185 #expect(throws: StatError.self) { try LeastSquares.levenbergMarquardt(residuals: { _ in [.nan] }, start: [1]) }186 #expect(throws: StatError.self) { try LeastSquares.curveFit(model: { x, _ in x }, x: [1, 2], y: [1], start: [1]) }187 #expect(throws: StatError.self) { try LeastSquares.curveFit(model: { x, _ in x }, x: [1, 2], y: [1, 2], start: [1], weights: [1]) }188 #expect(throws: StatError.self) {189 try LeastSquares.levenbergMarquardt(residuals: { p in [p[0], p[0] * 2] }, start: [1], jacobian: { _ in Matrix(rows: 1, cols: 1) })190 }191 }192}193