Files
gonum/optimize/unconstrained_test.go
Brendan Tracey d56ca496c1 optimize: Change Settings to allow InitialLocation (#497)
* optimize: Change Settings to allow InitialLocation

This modifies Settings to allow specifying an initial location and properties of the function (value, gradient, etc.). This allows to work with local optimizers that are seeded with initial settings. This has two fields that must be specified, InitX and InitValues. Ideally this would only be one location, but the difficulty is that the default value of the function is 0. We either must require the user to specify it is set (in this case that InitValues is non-zero), or require the user to change the default value away if it is not set. The former seems much safer.
2018-06-06 09:11:04 -06:00

1307 lines
31 KiB
Go

// Copyright ©2014 The Gonum Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package optimize
import (
"fmt"
"math"
"testing"
"gonum.org/v1/gonum/floats"
"gonum.org/v1/gonum/mat"
"gonum.org/v1/gonum/optimize/functions"
)
type unconstrainedTest struct {
// name is the name of the test.
name string
// p is the optimization problem to be solved.
p Problem
// x is the initial guess.
x []float64
// gradTol is the absolute gradient tolerance for the test. If gradTol == 0,
// the default value of 1e-12 will be used.
gradTol float64
// fAbsTol is the absolute function convergence for the test. If fAbsTol == 0,
// the default value of 1e-12 will be used.
fAbsTol float64
// fIter is the number of iterations for function convergence. If fIter == 0,
// the default value of 20 will be used.
fIter int
// long indicates that the test takes long time to finish and will be
// excluded if testing.Short returns true.
long bool
}
func (t unconstrainedTest) String() string {
dim := len(t.x)
if dim <= 10 {
// Print the initial X only for small-dimensional problems.
return fmt.Sprintf("F: %v\nDim: %v\nInitial X: %v\nGradientThreshold: %v",
t.name, dim, t.x, t.gradTol)
}
return fmt.Sprintf("F: %v\nDim: %v\nGradientThreshold: %v",
t.name, dim, t.gradTol)
}
var gradFreeTests = []unconstrainedTest{
{
name: "Beale",
p: Problem{
Func: functions.Beale{}.Func,
},
x: []float64{1, 1},
},
{
name: "BiggsEXP6",
p: Problem{
Func: functions.BiggsEXP6{}.Func,
},
x: []float64{1, 2, 1, 1, 1, 1},
},
{
name: "BrownAndDennis",
p: Problem{
Func: functions.BrownAndDennis{}.Func,
},
x: []float64{25, 5, -5, -1},
},
{
name: "ExtendedRosenbrock",
p: Problem{
Func: functions.ExtendedRosenbrock{}.Func,
},
x: []float64{-10, 10},
},
{
name: "ExtendedRosenbrock",
p: Problem{
Func: functions.ExtendedRosenbrock{}.Func,
},
x: []float64{-5, 4, 16, 3},
},
}
var gradientDescentTests = []unconstrainedTest{
{
name: "Beale",
p: Problem{
Func: functions.Beale{}.Func,
Grad: functions.Beale{}.Grad,
},
x: []float64{1, 1},
},
{
name: "Beale",
p: Problem{
Func: functions.Beale{}.Func,
Grad: functions.Beale{}.Grad,
},
x: []float64{3.00001, 0.50001},
},
{
name: "BiggsEXP2",
p: Problem{
Func: functions.BiggsEXP2{}.Func,
Grad: functions.BiggsEXP2{}.Grad,
},
x: []float64{1, 2},
},
{
name: "BiggsEXP2",
p: Problem{
Func: functions.BiggsEXP2{}.Func,
Grad: functions.BiggsEXP2{}.Grad,
},
x: []float64{1.00001, 10.00001},
},
{
name: "BiggsEXP3",
p: Problem{
Func: functions.BiggsEXP3{}.Func,
Grad: functions.BiggsEXP3{}.Grad,
},
x: []float64{1, 2, 1},
},
{
name: "BiggsEXP3",
p: Problem{
Func: functions.BiggsEXP3{}.Func,
Grad: functions.BiggsEXP3{}.Grad,
},
x: []float64{1.00001, 10.00001, 3.00001},
},
{
name: "ExtendedRosenbrock",
p: Problem{
Func: functions.ExtendedRosenbrock{}.Func,
Grad: functions.ExtendedRosenbrock{}.Grad,
},
x: []float64{-1.2, 1},
gradTol: 1e-10,
},
{
name: "ExtendedRosenbrock",
p: Problem{
Func: functions.ExtendedRosenbrock{}.Func,
Grad: functions.ExtendedRosenbrock{}.Grad,
},
x: []float64{1.00001, 1.00001},
gradTol: 1e-10,
},
{
name: "ExtendedRosenbrock",
p: Problem{
Func: functions.ExtendedRosenbrock{}.Func,
Grad: functions.ExtendedRosenbrock{}.Grad,
},
x: []float64{-1.2, 1, -1.2},
gradTol: 1e-10,
},
{
name: "ExtendedRosenbrock",
p: Problem{
Func: functions.ExtendedRosenbrock{}.Func,
Grad: functions.ExtendedRosenbrock{}.Grad,
},
x: []float64{-120, 100, 50},
long: true,
},
{
name: "ExtendedRosenbrock",
p: Problem{
Func: functions.ExtendedRosenbrock{}.Func,
Grad: functions.ExtendedRosenbrock{}.Grad,
},
x: []float64{1, 1, 1},
},
{
name: "ExtendedRosenbrock",
p: Problem{
Func: functions.ExtendedRosenbrock{}.Func,
Grad: functions.ExtendedRosenbrock{}.Grad,
},
x: []float64{1.00001, 1.00001, 1.00001},
gradTol: 1e-8,
},
{
name: "Gaussian",
p: Problem{
Func: functions.Gaussian{}.Func,
Grad: functions.Gaussian{}.Grad,
},
x: []float64{0.4, 1, 0},
gradTol: 1e-9,
},
{
name: "Gaussian",
p: Problem{
Func: functions.Gaussian{}.Func,
Grad: functions.Gaussian{}.Grad,
},
x: []float64{0.3989561, 1.0000191, 0},
gradTol: 1e-9,
},
{
name: "HelicalValley",
p: Problem{
Func: functions.HelicalValley{}.Func,
Grad: functions.HelicalValley{}.Grad,
},
x: []float64{-1, 0, 0},
},
{
name: "HelicalValley",
p: Problem{
Func: functions.HelicalValley{}.Func,
Grad: functions.HelicalValley{}.Grad,
},
x: []float64{1.00001, 0.00001, 0.00001},
},
{
name: "Trigonometric",
p: Problem{
Func: functions.Trigonometric{}.Func,
Grad: functions.Trigonometric{}.Grad,
},
x: []float64{0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1},
gradTol: 1e-7,
},
{
name: "Trigonometric",
p: Problem{
Func: functions.Trigonometric{}.Func,
Grad: functions.Trigonometric{}.Grad,
},
x: []float64{0.042964, 0.043976, 0.045093, 0.046338, 0.047744,
0.049354, 0.051237, 0.195209, 0.164977, 0.060148},
gradTol: 1e-8,
},
newVariablyDimensioned(2, 0),
{
name: "VariablyDimensioned",
p: Problem{
Func: functions.VariablyDimensioned{}.Func,
Grad: functions.VariablyDimensioned{}.Grad,
},
x: []float64{1.00001, 1.00001},
},
newVariablyDimensioned(10, 0),
{
name: "VariablyDimensioned",
p: Problem{
Func: functions.VariablyDimensioned{}.Func,
Grad: functions.VariablyDimensioned{}.Grad,
},
x: []float64{1.00001, 1.00001, 1.00001, 1.00001, 1.00001, 1.00001, 1.00001, 1.00001, 1.00001, 1.00001},
},
}
var cgTests = []unconstrainedTest{
{
name: "BiggsEXP4",
p: Problem{
Func: functions.BiggsEXP4{}.Func,
Grad: functions.BiggsEXP4{}.Grad,
},
x: []float64{1, 2, 1, 1},
},
{
name: "BiggsEXP4",
p: Problem{
Func: functions.BiggsEXP4{}.Func,
Grad: functions.BiggsEXP4{}.Grad,
},
x: []float64{1.00001, 10.00001, 1.00001, 5.00001},
},
{
name: "BiggsEXP5",
p: Problem{
Func: functions.BiggsEXP5{}.Func,
Grad: functions.BiggsEXP5{}.Grad,
},
x: []float64{1, 2, 1, 1, 1},
gradTol: 1e-7,
},
{
name: "BiggsEXP5",
p: Problem{
Func: functions.BiggsEXP5{}.Func,
Grad: functions.BiggsEXP5{}.Grad,
},
x: []float64{1.00001, 10.00001, 1.00001, 5.00001, 4.00001},
},
{
name: "BiggsEXP6",
p: Problem{
Func: functions.BiggsEXP6{}.Func,
Grad: functions.BiggsEXP6{}.Grad,
},
x: []float64{1, 2, 1, 1, 1, 1},
gradTol: 1e-7,
},
{
name: "BiggsEXP6",
p: Problem{
Func: functions.BiggsEXP6{}.Func,
Grad: functions.BiggsEXP6{}.Grad,
},
x: []float64{1.00001, 10.00001, 1.00001, 5.00001, 4.00001, 3.00001},
gradTol: 1e-8,
},
{
name: "Box3D",
p: Problem{
Func: functions.Box3D{}.Func,
Grad: functions.Box3D{}.Grad,
},
x: []float64{0, 10, 20},
},
{
name: "Box3D",
p: Problem{
Func: functions.Box3D{}.Func,
Grad: functions.Box3D{}.Grad,
},
x: []float64{1.00001, 10.00001, 1.00001},
},
{
name: "Box3D",
p: Problem{
Func: functions.Box3D{}.Func,
Grad: functions.Box3D{}.Grad,
},
x: []float64{100.00001, 100.00001, 0.00001},
},
{
name: "ExtendedPowellSingular",
p: Problem{
Func: functions.ExtendedPowellSingular{}.Func,
Grad: functions.ExtendedPowellSingular{}.Grad,
},
x: []float64{3, -1, 0, 3},
},
{
name: "ExtendedPowellSingular",
p: Problem{
Func: functions.ExtendedPowellSingular{}.Func,
Grad: functions.ExtendedPowellSingular{}.Grad,
},
x: []float64{0.00001, 0.00001, 0.00001, 0.00001},
},
{
name: "ExtendedPowellSingular",
p: Problem{
Func: functions.ExtendedPowellSingular{}.Func,
Grad: functions.ExtendedPowellSingular{}.Grad,
},
x: []float64{3, -1, 0, 3, 3, -1, 0, 3},
gradTol: 1e-8,
},
{
name: "ExtendedPowellSingular",
p: Problem{
Func: functions.ExtendedPowellSingular{}.Func,
Grad: functions.ExtendedPowellSingular{}.Grad,
},
x: []float64{0.00001, 0.00001, 0.00001, 0.00001, 0.00001, 0.00001, 0.00001, 0.00001},
},
{
name: "ExtendedRosenbrock",
p: Problem{
Func: functions.ExtendedRosenbrock{}.Func,
Grad: functions.ExtendedRosenbrock{}.Grad,
},
x: []float64{-1.2, 1, -1.2, 1},
},
{
name: "ExtendedRosenbrock",
p: Problem{
Func: functions.ExtendedRosenbrock{}.Func,
Grad: functions.ExtendedRosenbrock{}.Grad,
},
x: []float64{1e4, 1e4},
gradTol: 1e-10,
},
{
name: "ExtendedRosenbrock",
p: Problem{
Func: functions.ExtendedRosenbrock{}.Func,
Grad: functions.ExtendedRosenbrock{}.Grad,
},
x: []float64{1.00001, 1.00001, 1.00001, 1.00001},
gradTol: 1e-10,
},
{
name: "PenaltyI",
p: Problem{
Func: functions.PenaltyI{}.Func,
Grad: functions.PenaltyI{}.Grad,
},
x: []float64{1, 2, 3, 4, 5, 6, 7, 8, 9, 10},
gradTol: 1e-9,
},
{
name: "PenaltyI",
p: Problem{
Func: functions.PenaltyI{}.Func,
Grad: functions.PenaltyI{}.Grad,
},
x: []float64{0.250007, 0.250007, 0.250007, 0.250007},
gradTol: 1e-10,
},
{
name: "PenaltyI",
p: Problem{
Func: functions.PenaltyI{}.Func,
Grad: functions.PenaltyI{}.Grad,
},
x: []float64{0.1581, 0.1581, 0.1581, 0.1581, 0.1581, 0.1581,
0.1581, 0.1581, 0.1581, 0.1581},
gradTol: 1e-10,
},
{
name: "PenaltyII",
p: Problem{
Func: functions.PenaltyII{}.Func,
Grad: functions.PenaltyII{}.Grad,
},
x: []float64{0.5, 0.5, 0.5, 0.5},
gradTol: 1e-8,
},
{
name: "PenaltyII",
p: Problem{
Func: functions.PenaltyII{}.Func,
Grad: functions.PenaltyII{}.Grad,
},
x: []float64{0.19999, 0.19131, 0.4801, 0.51884},
gradTol: 1e-8,
},
{
name: "PenaltyII",
p: Problem{
Func: functions.PenaltyII{}.Func,
Grad: functions.PenaltyII{}.Grad,
},
x: []float64{0.19998, 0.01035, 0.01960, 0.03208, 0.04993, 0.07651,
0.11862, 0.19214, 0.34732, 0.36916},
gradTol: 1e-6,
},
{
name: "PowellBadlyScaled",
p: Problem{
Func: functions.PowellBadlyScaled{}.Func,
Grad: functions.PowellBadlyScaled{}.Grad,
},
x: []float64{1.09815e-05, 9.10614},
gradTol: 1e-8,
},
newVariablyDimensioned(100, 1e-10),
newVariablyDimensioned(1000, 1e-7),
newVariablyDimensioned(10000, 1e-4),
{
name: "Watson",
p: Problem{
Func: functions.Watson{}.Func,
Grad: functions.Watson{}.Grad,
},
x: []float64{0, 0, 0, 0, 0, 0},
gradTol: 1e-6,
},
{
name: "Watson",
p: Problem{
Func: functions.Watson{}.Func,
Grad: functions.Watson{}.Grad,
},
x: []float64{-0.01572, 1.01243, -0.23299, 1.26043, -1.51372, 0.99299},
gradTol: 1e-6,
},
{
name: "Watson",
p: Problem{
Func: functions.Watson{}.Func,
Grad: functions.Watson{}.Grad,
},
x: []float64{0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
gradTol: 1e-6,
long: true,
},
{
name: "Watson",
p: Problem{
Func: functions.Watson{}.Func,
Grad: functions.Watson{}.Grad,
},
x: []float64{-1.53070e-05, 0.99978, 0.01476, 0.14634, 1.00082,
-2.61773, 4.10440, -3.14361, 1.05262},
gradTol: 1e-6,
},
{
name: "Wood",
p: Problem{
Func: functions.Wood{}.Func,
Grad: functions.Wood{}.Grad,
},
x: []float64{-3, -1, -3, -1},
gradTol: 1e-6,
},
}
var quasiNewtonTests = []unconstrainedTest{
{
name: "BiggsEXP4",
p: Problem{
Func: functions.BiggsEXP4{}.Func,
Grad: functions.BiggsEXP4{}.Grad,
},
x: []float64{1, 2, 1, 1},
},
{
name: "BiggsEXP4",
p: Problem{
Func: functions.BiggsEXP4{}.Func,
Grad: functions.BiggsEXP4{}.Grad,
},
x: []float64{1.00001, 10.00001, 1.00001, 5.00001},
},
{
name: "BiggsEXP5",
p: Problem{
Func: functions.BiggsEXP5{}.Func,
Grad: functions.BiggsEXP5{}.Grad,
},
x: []float64{1, 2, 1, 1, 1},
gradTol: 1e-10,
},
{
name: "BiggsEXP5",
p: Problem{
Func: functions.BiggsEXP5{}.Func,
Grad: functions.BiggsEXP5{}.Grad,
},
x: []float64{1.00001, 10.00001, 1.00001, 5.00001, 4.00001},
},
{
name: "BiggsEXP6",
p: Problem{
Func: functions.BiggsEXP6{}.Func,
Grad: functions.BiggsEXP6{}.Grad,
},
x: []float64{1, 2, 1, 1, 1, 1},
gradTol: 1e-8,
},
{
name: "BiggsEXP6",
p: Problem{
Func: functions.BiggsEXP6{}.Func,
Grad: functions.BiggsEXP6{}.Grad,
},
x: []float64{1.00001, 10.00001, 1.00001, 5.00001, 4.00001, 3.00001},
gradTol: 1e-8,
},
{
name: "Box3D",
p: Problem{
Func: functions.Box3D{}.Func,
Grad: functions.Box3D{}.Grad,
},
x: []float64{0, 10, 20},
},
{
name: "Box3D",
p: Problem{
Func: functions.Box3D{}.Func,
Grad: functions.Box3D{}.Grad,
},
x: []float64{1.00001, 10.00001, 1.00001},
},
{
name: "Box3D",
p: Problem{
Func: functions.Box3D{}.Func,
Grad: functions.Box3D{}.Grad,
},
x: []float64{100.00001, 100.00001, 0.00001},
},
{
name: "BrownBadlyScaled",
p: Problem{
Func: functions.BrownBadlyScaled{}.Func,
Grad: functions.BrownBadlyScaled{}.Grad,
},
x: []float64{1, 1},
},
{
name: "BrownBadlyScaled",
p: Problem{
Func: functions.BrownBadlyScaled{}.Func,
Grad: functions.BrownBadlyScaled{}.Grad,
},
x: []float64{1.000001e6, 2.01e-6},
},
{
name: "ExtendedPowellSingular",
p: Problem{
Func: functions.ExtendedPowellSingular{}.Func,
Grad: functions.ExtendedPowellSingular{}.Grad,
},
x: []float64{3, -1, 0, 3},
},
{
name: "ExtendedPowellSingular",
p: Problem{
Func: functions.ExtendedPowellSingular{}.Func,
Grad: functions.ExtendedPowellSingular{}.Grad,
},
x: []float64{0.00001, 0.00001, 0.00001, 0.00001},
},
{
name: "ExtendedPowellSingular",
p: Problem{
Func: functions.ExtendedPowellSingular{}.Func,
Grad: functions.ExtendedPowellSingular{}.Grad,
},
x: []float64{3, -1, 0, 3, 3, -1, 0, 3},
},
{
name: "ExtendedPowellSingular",
p: Problem{
Func: functions.ExtendedPowellSingular{}.Func,
Grad: functions.ExtendedPowellSingular{}.Grad,
},
x: []float64{0.00001, 0.00001, 0.00001, 0.00001, 0.00001, 0.00001, 0.00001, 0.00001},
},
{
name: "ExtendedRosenbrock",
p: Problem{
Func: functions.ExtendedRosenbrock{}.Func,
Grad: functions.ExtendedRosenbrock{}.Grad,
},
x: []float64{-1.2, 1, -1.2, 1},
},
{
name: "ExtendedRosenbrock",
p: Problem{
Func: functions.ExtendedRosenbrock{}.Func,
Grad: functions.ExtendedRosenbrock{}.Grad,
},
x: []float64{1.00001, 1.00001, 1.00001, 1.00001},
},
{
name: "Gaussian",
p: Problem{
Func: functions.Gaussian{}.Func,
Grad: functions.Gaussian{}.Grad,
},
x: []float64{0.4, 1, 0},
gradTol: 1e-11,
},
{
name: "GulfResearchAndDevelopment",
p: Problem{
Func: functions.GulfResearchAndDevelopment{}.Func,
Grad: functions.GulfResearchAndDevelopment{}.Grad,
},
x: []float64{5, 2.5, 0.15},
},
{
name: "GulfResearchAndDevelopment",
p: Problem{
Func: functions.GulfResearchAndDevelopment{}.Func,
Grad: functions.GulfResearchAndDevelopment{}.Grad,
},
x: []float64{50.00001, 25.00001, 1.50001},
},
{
name: "GulfResearchAndDevelopment",
p: Problem{
Func: functions.GulfResearchAndDevelopment{}.Func,
Grad: functions.GulfResearchAndDevelopment{}.Grad,
},
x: []float64{99.89529, 60.61453, 9.16124},
},
{
name: "GulfResearchAndDevelopment",
p: Problem{
Func: functions.GulfResearchAndDevelopment{}.Func,
Grad: functions.GulfResearchAndDevelopment{}.Grad,
},
x: []float64{201.66258, 60.61633, 10.22489},
},
{
name: "PenaltyI",
p: Problem{
Func: functions.PenaltyI{}.Func,
Grad: functions.PenaltyI{}.Grad,
},
x: []float64{1, 2, 3, 4, 5, 6, 7, 8, 9, 10},
},
{
name: "PenaltyI",
p: Problem{
Func: functions.PenaltyI{}.Func,
Grad: functions.PenaltyI{}.Grad,
},
x: []float64{0.250007, 0.250007, 0.250007, 0.250007},
gradTol: 1e-9,
},
{
name: "PenaltyI",
p: Problem{
Func: functions.PenaltyI{}.Func,
Grad: functions.PenaltyI{}.Grad,
},
x: []float64{0.1581, 0.1581, 0.1581, 0.1581, 0.1581, 0.1581,
0.1581, 0.1581, 0.1581, 0.1581},
},
{
name: "PenaltyII",
p: Problem{
Func: functions.PenaltyII{}.Func,
Grad: functions.PenaltyII{}.Grad,
},
x: []float64{0.5, 0.5, 0.5, 0.5},
gradTol: 1e-10,
},
{
name: "PenaltyII",
p: Problem{
Func: functions.PenaltyII{}.Func,
Grad: functions.PenaltyII{}.Grad,
},
x: []float64{0.19999, 0.19131, 0.4801, 0.51884},
gradTol: 1e-10,
},
{
name: "PenaltyII",
p: Problem{
Func: functions.PenaltyII{}.Func,
Grad: functions.PenaltyII{}.Grad,
},
x: []float64{0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5},
gradTol: 1e-9,
},
{
name: "PenaltyII",
p: Problem{
Func: functions.PenaltyII{}.Func,
Grad: functions.PenaltyII{}.Grad,
},
x: []float64{0.19998, 0.01035, 0.01960, 0.03208, 0.04993, 0.07651,
0.11862, 0.19214, 0.34732, 0.36916},
gradTol: 1e-9,
},
{
name: "PowellBadlyScaled",
p: Problem{
Func: functions.PowellBadlyScaled{}.Func,
Grad: functions.PowellBadlyScaled{}.Grad,
},
x: []float64{0, 1},
},
{
name: "PowellBadlyScaled",
p: Problem{
Func: functions.PowellBadlyScaled{}.Func,
Grad: functions.PowellBadlyScaled{}.Grad,
},
x: []float64{1.09815e-05, 9.10614},
gradTol: 1e-10,
},
newVariablyDimensioned(100, 1e-10),
{
name: "Watson",
p: Problem{
Func: functions.Watson{}.Func,
Grad: functions.Watson{}.Grad,
},
x: []float64{0, 0, 0, 0, 0, 0},
gradTol: 1e-7,
},
{
name: "Watson",
p: Problem{
Func: functions.Watson{}.Func,
Grad: functions.Watson{}.Grad,
},
x: []float64{-0.01572, 1.01243, -0.23299, 1.26043, -1.51372, 0.99299},
gradTol: 1e-7,
},
{
name: "Watson",
p: Problem{
Func: functions.Watson{}.Func,
Grad: functions.Watson{}.Grad,
},
x: []float64{0, 0, 0, 0, 0, 0, 0, 0, 0},
gradTol: 1e-8,
},
{
name: "Watson",
p: Problem{
Func: functions.Watson{}.Func,
Grad: functions.Watson{}.Grad,
},
x: []float64{-1.53070e-05, 0.99978, 0.01476, 0.14634, 1.00082,
-2.61773, 4.10440, -3.14361, 1.05262},
gradTol: 1e-8,
},
}
var bfgsTests = []unconstrainedTest{
{
name: "BiggsEXP6",
p: Problem{
Func: functions.BiggsEXP6{}.Func,
Grad: functions.BiggsEXP6{}.Grad,
},
x: []float64{1, 2, 1, 1, 1, 1},
gradTol: 1e-10,
},
{
name: "BiggsEXP6",
p: Problem{
Func: functions.BiggsEXP6{}.Func,
Grad: functions.BiggsEXP6{}.Grad,
},
x: []float64{1.00001, 10.00001, 1.00001, 5.00001, 4.00001, 3.00001},
gradTol: 1e-10,
},
{
name: "BrownAndDennis",
p: Problem{
Func: functions.BrownAndDennis{}.Func,
Grad: functions.BrownAndDennis{}.Grad,
},
x: []float64{25, 5, -5, -1},
gradTol: 1e-5,
},
{
name: "ExtendedRosenbrock",
p: Problem{
Func: functions.ExtendedRosenbrock{}.Func,
Grad: functions.ExtendedRosenbrock{}.Grad,
},
x: []float64{1e5, 1e5},
gradTol: 1e-10,
},
{
name: "Gaussian",
p: Problem{
Func: functions.Gaussian{}.Func,
Grad: functions.Gaussian{}.Grad,
},
x: []float64{0.398, 1, 0},
gradTol: 1e-11,
},
{
name: "Wood",
p: Problem{
Func: functions.Wood{}.Func,
Grad: functions.Wood{}.Grad,
},
x: []float64{-3, -1, -3, -1},
},
}
var lbfgsTests = []unconstrainedTest{
{
name: "BiggsEXP6",
p: Problem{
Func: functions.BiggsEXP6{}.Func,
Grad: functions.BiggsEXP6{}.Grad,
},
x: []float64{1, 2, 1, 1, 1, 1},
gradTol: 1e-8,
},
{
name: "BiggsEXP6",
p: Problem{
Func: functions.BiggsEXP6{}.Func,
Grad: functions.BiggsEXP6{}.Grad,
},
x: []float64{1.00001, 10.00001, 1.00001, 5.00001, 4.00001, 3.00001},
gradTol: 1e-8,
},
{
name: "ExtendedRosenbrock",
p: Problem{
Func: functions.ExtendedRosenbrock{}.Func,
Grad: functions.ExtendedRosenbrock{}.Grad,
},
x: []float64{1e7, 1e6},
gradTol: 1e-10,
},
{
name: "Gaussian",
p: Problem{
Func: functions.Gaussian{}.Func,
Grad: functions.Gaussian{}.Grad,
},
x: []float64{0.398, 1, 0},
gradTol: 1e-10,
},
newVariablyDimensioned(1000, 1e-8),
newVariablyDimensioned(10000, 1e-5),
}
var newtonTests = []unconstrainedTest{
{
name: "Beale",
p: Problem{
Func: functions.Beale{}.Func,
Grad: functions.Beale{}.Grad,
Hess: functions.Beale{}.Hess,
},
x: []float64{1, 1},
},
{
name: "BrownAndDennis",
p: Problem{
Func: functions.BrownAndDennis{}.Func,
Grad: functions.BrownAndDennis{}.Grad,
Hess: functions.BrownAndDennis{}.Hess,
},
x: []float64{25, 5, -5, -1},
gradTol: 1e-10,
},
{
name: "BrownBadlyScaled",
p: Problem{
Func: functions.BrownBadlyScaled{}.Func,
Grad: functions.BrownBadlyScaled{}.Grad,
Hess: functions.BrownBadlyScaled{}.Hess,
},
x: []float64{1, 1},
},
{
name: "PowellBadlyScaled",
p: Problem{
Func: functions.PowellBadlyScaled{}.Func,
Grad: functions.PowellBadlyScaled{}.Grad,
Hess: functions.PowellBadlyScaled{}.Hess,
},
x: []float64{0, 1},
gradTol: 1e-10,
},
{
name: "Watson",
p: Problem{
Func: functions.Watson{}.Func,
Grad: functions.Watson{}.Grad,
Hess: functions.Watson{}.Hess,
},
x: []float64{0, 0, 0, 0, 0, 0},
},
{
name: "Watson",
p: Problem{
Func: functions.Watson{}.Func,
Grad: functions.Watson{}.Grad,
Hess: functions.Watson{}.Hess,
},
x: []float64{0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
},
{
name: "Wood",
p: Problem{
Func: functions.Wood{}.Func,
Grad: functions.Wood{}.Grad,
Hess: functions.Wood{}.Hess,
},
x: []float64{-3, -1, -3, -1},
},
}
func newVariablyDimensioned(dim int, gradTol float64) unconstrainedTest {
x := make([]float64, dim)
for i := range x {
x[i] = float64(dim-i-1) / float64(dim)
}
return unconstrainedTest{
name: "VariablyDimensioned",
p: Problem{
Func: functions.VariablyDimensioned{}.Func,
Grad: functions.VariablyDimensioned{}.Grad,
},
x: x,
gradTol: gradTol,
}
}
func TestLocal(t *testing.T) {
var tests []unconstrainedTest
// Mix of functions with and without Grad method.
tests = append(tests, gradFreeTests...)
tests = append(tests, gradientDescentTests...)
testLocal(t, tests, nil)
}
func TestNelderMead(t *testing.T) {
var tests []unconstrainedTest
// Mix of functions with and without Grad method.
tests = append(tests, gradFreeTests...)
tests = append(tests, gradientDescentTests...)
testLocal(t, tests, &NelderMead{})
}
func TestGradientDescent(t *testing.T) {
testLocal(t, gradientDescentTests, &GradientDescent{})
}
func TestGradientDescentBacktracking(t *testing.T) {
testLocal(t, gradientDescentTests, &GradientDescent{
Linesearcher: &Backtracking{
DecreaseFactor: 0.1,
},
})
}
func TestGradientDescentBisection(t *testing.T) {
testLocal(t, gradientDescentTests, &GradientDescent{
Linesearcher: &Bisection{},
})
}
func TestCG(t *testing.T) {
var tests []unconstrainedTest
tests = append(tests, gradientDescentTests...)
tests = append(tests, cgTests...)
testLocal(t, tests, &CG{})
}
func TestFletcherReevesQuadStep(t *testing.T) {
var tests []unconstrainedTest
tests = append(tests, gradientDescentTests...)
tests = append(tests, cgTests...)
testLocal(t, tests, &CG{
Variant: &FletcherReeves{},
InitialStep: &QuadraticStepSize{},
})
}
func TestFletcherReevesFirstOrderStep(t *testing.T) {
var tests []unconstrainedTest
tests = append(tests, gradientDescentTests...)
tests = append(tests, cgTests...)
testLocal(t, tests, &CG{
Variant: &FletcherReeves{},
InitialStep: &FirstOrderStepSize{},
})
}
func TestHestenesStiefelQuadStep(t *testing.T) {
var tests []unconstrainedTest
tests = append(tests, gradientDescentTests...)
tests = append(tests, cgTests...)
testLocal(t, tests, &CG{
Variant: &HestenesStiefel{},
InitialStep: &QuadraticStepSize{},
})
}
func TestHestenesStiefelFirstOrderStep(t *testing.T) {
var tests []unconstrainedTest
tests = append(tests, gradientDescentTests...)
tests = append(tests, cgTests...)
testLocal(t, tests, &CG{
Variant: &HestenesStiefel{},
InitialStep: &FirstOrderStepSize{},
})
}
func TestPolakRibiereQuadStep(t *testing.T) {
var tests []unconstrainedTest
tests = append(tests, gradientDescentTests...)
tests = append(tests, cgTests...)
testLocal(t, tests, &CG{
Variant: &PolakRibierePolyak{},
InitialStep: &QuadraticStepSize{},
})
}
func TestPolakRibiereFirstOrderStep(t *testing.T) {
var tests []unconstrainedTest
tests = append(tests, gradientDescentTests...)
tests = append(tests, cgTests...)
testLocal(t, tests, &CG{
Variant: &PolakRibierePolyak{},
InitialStep: &FirstOrderStepSize{},
})
}
func TestDaiYuanQuadStep(t *testing.T) {
var tests []unconstrainedTest
tests = append(tests, gradientDescentTests...)
tests = append(tests, cgTests...)
testLocal(t, tests, &CG{
Variant: &DaiYuan{},
InitialStep: &QuadraticStepSize{},
})
}
func TestDaiYuanFirstOrderStep(t *testing.T) {
var tests []unconstrainedTest
tests = append(tests, gradientDescentTests...)
tests = append(tests, cgTests...)
testLocal(t, tests, &CG{
Variant: &DaiYuan{},
InitialStep: &FirstOrderStepSize{},
})
}
func TestHagerZhangQuadStep(t *testing.T) {
var tests []unconstrainedTest
tests = append(tests, gradientDescentTests...)
tests = append(tests, cgTests...)
testLocal(t, tests, &CG{
Variant: &HagerZhang{},
InitialStep: &QuadraticStepSize{},
})
}
func TestHagerZhangFirstOrderStep(t *testing.T) {
var tests []unconstrainedTest
tests = append(tests, gradientDescentTests...)
tests = append(tests, cgTests...)
testLocal(t, tests, &CG{
Variant: &HagerZhang{},
InitialStep: &FirstOrderStepSize{},
})
}
func TestBFGS(t *testing.T) {
var tests []unconstrainedTest
tests = append(tests, gradientDescentTests...)
tests = append(tests, quasiNewtonTests...)
tests = append(tests, bfgsTests...)
testLocal(t, tests, &BFGS{})
}
func TestLBFGS(t *testing.T) {
var tests []unconstrainedTest
tests = append(tests, gradientDescentTests...)
tests = append(tests, quasiNewtonTests...)
tests = append(tests, lbfgsTests...)
testLocal(t, tests, &LBFGS{})
}
func TestNewton(t *testing.T) {
testLocal(t, newtonTests, &Newton{})
}
func testLocal(t *testing.T, tests []unconstrainedTest, method Method) {
for cas, test := range tests {
if test.long && testing.Short() {
continue
}
settings := DefaultSettings()
settings.Recorder = nil
if method != nil && method.Needs().Gradient {
// Turn off function convergence checks for gradient-based methods.
settings.FunctionConverge = nil
} else {
if test.fIter == 0 {
test.fIter = 20
}
settings.FunctionConverge.Iterations = test.fIter
if test.fAbsTol == 0 {
test.fAbsTol = 1e-12
}
settings.FunctionConverge.Absolute = test.fAbsTol
}
if test.gradTol == 0 {
test.gradTol = 1e-12
}
settings.GradientThreshold = test.gradTol
result, err := Local(test.p, test.x, settings, method)
if err != nil {
t.Errorf("Case %d: error finding minimum (%v) for:\n%v", cas, err, test)
continue
}
if result == nil {
t.Errorf("Case %d: nil result without error for:\n%v", cas, test)
continue
}
// Check that the function value at the found optimum location is
// equal to result.F.
optF := test.p.Func(result.X)
if optF != result.F {
t.Errorf("Case %d: Function value at the optimum location %v not equal to the returned value %v for:\n%v",
cas, optF, result.F, test)
}
if result.Gradient != nil {
// Evaluate the norm of the gradient at the found optimum location.
g := make([]float64, len(test.x))
test.p.Grad(g, result.X)
if !floats.Equal(result.Gradient, g) {
t.Errorf("Case %d: Gradient at the optimum location not equal to the returned value for:\n%v", cas, test)
}
optNorm := floats.Norm(g, math.Inf(1))
// Check that the norm of the gradient at the found optimum location is
// smaller than the tolerance.
if optNorm >= settings.GradientThreshold {
t.Errorf("Case %d: Norm of the gradient at the optimum location %v not smaller than tolerance %v for:\n%v",
cas, optNorm, settings.GradientThreshold, test)
}
}
if method == nil {
// The tests below make sense only if the method used is known.
continue
}
if !method.Needs().Gradient && !method.Needs().Hessian {
// Gradient-free tests can correctly terminate only with
// FunctionConvergence status.
if result.Status != FunctionConvergence {
t.Errorf("Status not %v, %v instead", FunctionConvergence, result.Status)
}
}
// We are going to restart the solution using known initial data, so
// evaluate them.
settings.InitX = test.x
settings.InitValues = &Location{}
settings.InitValues.F = test.p.Func(test.x)
if method.Needs().Gradient {
settings.InitValues.Gradient = resize(settings.InitValues.Gradient, len(test.x))
test.p.Grad(settings.InitValues.Gradient, test.x)
}
if method.Needs().Hessian {
settings.InitValues.Hessian = mat.NewSymDense(len(test.x), nil)
test.p.Hess(settings.InitValues.Hessian, test.x)
}
// Rerun the test again to make sure that it gets the same answer with
// the same starting condition. Moreover, we are using the initial data.
result2, err2 := Local(test.p, test.x, settings, method)
if err2 != nil {
t.Errorf("error finding minimum second time (%v) for:\n%v", err2, test)
continue
}
if result2 == nil {
t.Errorf("second time nil result without error for:\n%v", test)
continue
}
// At the moment all the optimizers are deterministic, so check that we
// get _exactly_ the same answer second time as well.
if result.F != result2.F || !floats.Equal(result.X, result2.X) {
t.Errorf("Different minimum second time for:\n%v", test)
}
// Check that providing initial data reduces the number of evaluations exactly by one.
if result.FuncEvaluations != result2.FuncEvaluations+1 {
t.Errorf("Providing initial data does not reduce the number of Func calls for:\n%v", test)
continue
}
if method.Needs().Gradient {
if result.GradEvaluations != result2.GradEvaluations+1 {
t.Errorf("Providing initial data does not reduce the number of Grad calls for:\n%v", test)
continue
}
}
if method.Needs().Hessian {
if result.HessEvaluations != result2.HessEvaluations+1 {
t.Errorf("Providing initial data does not reduce the number of Hess calls for:\n%v", test)
continue
}
}
}
}
func TestIssue76(t *testing.T) {
p := Problem{
Func: functions.BrownAndDennis{}.Func,
Grad: functions.BrownAndDennis{}.Grad,
}
// Location very close to the minimum.
x := []float64{-11.594439904886773, 13.203630051265385, -0.40343948776868443, 0.2367787746745986}
s := &Settings{
FunctionThreshold: math.Inf(-1),
GradientThreshold: 1e-14,
MajorIterations: 1000000,
}
m := &GradientDescent{
Linesearcher: &Backtracking{},
}
// We are not interested in the error, only in the returned status.
r, _ := Local(p, x, s, m)
// With the above stringent tolerance, the optimizer will never
// successfully reach the minimum. Check if it terminated in a finite
// number of steps.
if r.Status == IterationLimit {
t.Error("Issue https://github.com/gonum/optimize/issues/76 not fixed")
}
}