blob: dca67bacfdab07e5a37ec5e170ef531741040fcc [file] [log] [blame]
Austin Schuh70cc9552019-01-21 19:46:48 -08001// Ceres Solver - A fast non-linear least squares minimizer
Austin Schuh3de38b02024-06-25 18:25:10 -07002// Copyright 2024 Google Inc. All rights reserved.
Austin Schuh70cc9552019-01-21 19:46:48 -08003// http://ceres-solver.org/
4//
5// Redistribution and use in source and binary forms, with or without
6// modification, are permitted provided that the following conditions are met:
7//
8// * Redistributions of source code must retain the above copyright notice,
9// this list of conditions and the following disclaimer.
10// * Redistributions in binary form must reproduce the above copyright notice,
11// this list of conditions and the following disclaimer in the documentation
12// and/or other materials provided with the distribution.
13// * Neither the name of Google Inc. nor the names of its contributors may be
14// used to endorse or promote products derived from this software without
15// specific prior written permission.
16//
17// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
18// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20// ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
21// LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
22// CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
23// SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
24// INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
25// CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
26// ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
27// POSSIBILITY OF SUCH DAMAGE.
28//
29// Author: sameeragarwal@google.com (Sameer Agarwal)
30
31#include "ceres/autodiff_cost_function.h"
32
Austin Schuh70cc9552019-01-21 19:46:48 -080033#include <memory>
34
Austin Schuh70cc9552019-01-21 19:46:48 -080035#include "ceres/array_utils.h"
Austin Schuh1d1e6ea2020-12-23 21:56:30 -080036#include "ceres/cost_function.h"
37#include "gtest/gtest.h"
Austin Schuh70cc9552019-01-21 19:46:48 -080038
Austin Schuh3de38b02024-06-25 18:25:10 -070039namespace ceres::internal {
Austin Schuh70cc9552019-01-21 19:46:48 -080040
41class BinaryScalarCost {
42 public:
Austin Schuh1d1e6ea2020-12-23 21:56:30 -080043 explicit BinaryScalarCost(double a) : a_(a) {}
Austin Schuh70cc9552019-01-21 19:46:48 -080044 template <typename T>
Austin Schuh1d1e6ea2020-12-23 21:56:30 -080045 bool operator()(const T* const x, const T* const y, T* cost) const {
46 cost[0] = x[0] * y[0] + x[1] * y[1] - T(a_);
Austin Schuh70cc9552019-01-21 19:46:48 -080047 return true;
48 }
Austin Schuh1d1e6ea2020-12-23 21:56:30 -080049
Austin Schuh70cc9552019-01-21 19:46:48 -080050 private:
51 double a_;
52};
53
54TEST(AutodiffCostFunction, BilinearDifferentiationTest) {
Austin Schuh1d1e6ea2020-12-23 21:56:30 -080055 CostFunction* cost_function =
56 new AutoDiffCostFunction<BinaryScalarCost, 1, 2, 2>(
57 new BinaryScalarCost(1.0));
Austin Schuh70cc9552019-01-21 19:46:48 -080058
Austin Schuh3de38b02024-06-25 18:25:10 -070059 auto** parameters = new double*[2];
Austin Schuh70cc9552019-01-21 19:46:48 -080060 parameters[0] = new double[2];
61 parameters[1] = new double[2];
62
63 parameters[0][0] = 1;
64 parameters[0][1] = 2;
65
66 parameters[1][0] = 3;
67 parameters[1][1] = 4;
68
Austin Schuh3de38b02024-06-25 18:25:10 -070069 auto** jacobians = new double*[2];
Austin Schuh70cc9552019-01-21 19:46:48 -080070 jacobians[0] = new double[2];
71 jacobians[1] = new double[2];
72
73 double residuals = 0.0;
74
Austin Schuh1d1e6ea2020-12-23 21:56:30 -080075 cost_function->Evaluate(parameters, &residuals, nullptr);
Austin Schuh70cc9552019-01-21 19:46:48 -080076 EXPECT_EQ(10.0, residuals);
77
78 cost_function->Evaluate(parameters, &residuals, jacobians);
79 EXPECT_EQ(10.0, residuals);
80
81 EXPECT_EQ(3, jacobians[0][0]);
82 EXPECT_EQ(4, jacobians[0][1]);
83 EXPECT_EQ(1, jacobians[1][0]);
84 EXPECT_EQ(2, jacobians[1][1]);
85
86 delete[] jacobians[0];
87 delete[] jacobians[1];
88 delete[] parameters[0];
89 delete[] parameters[1];
90 delete[] jacobians;
91 delete[] parameters;
92 delete cost_function;
93}
94
95struct TenParameterCost {
96 template <typename T>
97 bool operator()(const T* const x0,
98 const T* const x1,
99 const T* const x2,
100 const T* const x3,
101 const T* const x4,
102 const T* const x5,
103 const T* const x6,
104 const T* const x7,
105 const T* const x8,
106 const T* const x9,
107 T* cost) const {
108 cost[0] = *x0 + *x1 + *x2 + *x3 + *x4 + *x5 + *x6 + *x7 + *x8 + *x9;
109 return true;
110 }
111};
112
113TEST(AutodiffCostFunction, ManyParameterAutodiffInstantiates) {
Austin Schuh1d1e6ea2020-12-23 21:56:30 -0800114 CostFunction* cost_function =
115 new AutoDiffCostFunction<TenParameterCost,
116 1,
117 1,
118 1,
119 1,
120 1,
121 1,
122 1,
123 1,
124 1,
125 1,
126 1>(new TenParameterCost);
Austin Schuh70cc9552019-01-21 19:46:48 -0800127
Austin Schuh3de38b02024-06-25 18:25:10 -0700128 auto** parameters = new double*[10];
129 auto** jacobians = new double*[10];
Austin Schuh70cc9552019-01-21 19:46:48 -0800130 for (int i = 0; i < 10; ++i) {
131 parameters[i] = new double[1];
132 parameters[i][0] = i;
133 jacobians[i] = new double[1];
134 }
135
136 double residuals = 0.0;
137
Austin Schuh1d1e6ea2020-12-23 21:56:30 -0800138 cost_function->Evaluate(parameters, &residuals, nullptr);
Austin Schuh70cc9552019-01-21 19:46:48 -0800139 EXPECT_EQ(45.0, residuals);
140
141 cost_function->Evaluate(parameters, &residuals, jacobians);
142 EXPECT_EQ(residuals, 45.0);
143 for (int i = 0; i < 10; ++i) {
144 EXPECT_EQ(1.0, jacobians[i][0]);
145 }
146
147 for (int i = 0; i < 10; ++i) {
148 delete[] jacobians[i];
149 delete[] parameters[i];
150 }
151 delete[] jacobians;
152 delete[] parameters;
153 delete cost_function;
154}
155
156struct OnlyFillsOneOutputFunctor {
157 template <typename T>
158 bool operator()(const T* x, T* output) const {
159 output[0] = x[0];
160 return true;
161 }
162};
163
164TEST(AutoDiffCostFunction, PartiallyFilledResidualShouldFailEvaluation) {
165 double parameter = 1.0;
166 double jacobian[2];
167 double residuals[2];
168 double* parameters[] = {&parameter};
169 double* jacobians[] = {jacobian};
170
171 std::unique_ptr<CostFunction> cost_function(
172 new AutoDiffCostFunction<OnlyFillsOneOutputFunctor, 2, 1>(
173 new OnlyFillsOneOutputFunctor));
174 InvalidateArray(2, jacobian);
175 InvalidateArray(2, residuals);
176 EXPECT_TRUE(cost_function->Evaluate(parameters, residuals, jacobians));
177 EXPECT_FALSE(IsArrayValid(2, jacobian));
178 EXPECT_FALSE(IsArrayValid(2, residuals));
179}
180
Austin Schuh3de38b02024-06-25 18:25:10 -0700181TEST(AutodiffCostFunction, ArgumentForwarding) {
182 // No narrowing conversion warning should be emitted
183 auto cost_function1 =
184 std::make_unique<AutoDiffCostFunction<BinaryScalarCost, 1, 2, 2>>(1);
185 auto cost_function2 =
186 std::make_unique<AutoDiffCostFunction<BinaryScalarCost, 1, 2, 2>>(2.0);
187 // Default constructible functor
188 auto cost_function3 =
189 std::make_unique<AutoDiffCostFunction<OnlyFillsOneOutputFunctor, 1, 1>>();
190}
191
192TEST(AutodiffCostFunction, UniquePtrCtor) {
193 auto cost_function1 =
194 std::make_unique<AutoDiffCostFunction<BinaryScalarCost, 1, 2, 2>>(
195 std::make_unique<BinaryScalarCost>(1));
196 auto cost_function2 =
197 std::make_unique<AutoDiffCostFunction<BinaryScalarCost, 1, 2, 2>>(
198 std::make_unique<BinaryScalarCost>(2.0));
199}
200
201} // namespace ceres::internal