Home | History | Annotate | Download | only in eigen2
      1 // This file is part of Eigen, a lightweight C++ template library
      2 // for linear algebra. Eigen itself is part of the KDE project.
      3 //
      4 // Copyright (C) 2008 Benoit Jacob <jacob.benoit.1 (at) gmail.com>
      5 //
      6 // This Source Code Form is subject to the terms of the Mozilla
      7 // Public License v. 2.0. If a copy of the MPL was not distributed
      8 // with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
      9 
     10 #include "main.h"
     11 #include <Eigen/LeastSquares>
     12 
     13 template<typename VectorType,
     14          typename HyperplaneType>
     15 void makeNoisyCohyperplanarPoints(int numPoints,
     16                                   VectorType **points,
     17                                   HyperplaneType *hyperplane,
     18                                   typename VectorType::Scalar noiseAmplitude)
     19 {
     20   typedef typename VectorType::Scalar Scalar;
     21   const int size = points[0]->size();
     22   // pick a random hyperplane, store the coefficients of its equation
     23   hyperplane->coeffs().resize(size + 1);
     24   for(int j = 0; j < size + 1; j++)
     25   {
     26     do {
     27       hyperplane->coeffs().coeffRef(j) = ei_random<Scalar>();
     28     } while(ei_abs(hyperplane->coeffs().coeff(j)) < 0.5);
     29   }
     30 
     31   // now pick numPoints random points on this hyperplane
     32   for(int i = 0; i < numPoints; i++)
     33   {
     34     VectorType& cur_point = *(points[i]);
     35     do
     36     {
     37       cur_point = VectorType::Random(size)/*.normalized()*/;
     38       // project cur_point onto the hyperplane
     39       Scalar x = - (hyperplane->coeffs().start(size).cwise()*cur_point).sum();
     40       cur_point *= hyperplane->coeffs().coeff(size) / x;
     41     } while( cur_point.norm() < 0.5
     42           || cur_point.norm() > 2.0 );
     43   }
     44 
     45   // add some noise to these points
     46   for(int i = 0; i < numPoints; i++ )
     47     *(points[i]) += noiseAmplitude * VectorType::Random(size);
     48 }
     49 
     50 template<typename VectorType>
     51 void check_linearRegression(int numPoints,
     52                             VectorType **points,
     53                             const VectorType& original,
     54                             typename VectorType::Scalar tolerance)
     55 {
     56   int size = points[0]->size();
     57   assert(size==2);
     58   VectorType result(size);
     59   linearRegression(numPoints, points, &result, 1);
     60   typename VectorType::Scalar error = (result - original).norm() / original.norm();
     61   VERIFY(ei_abs(error) < ei_abs(tolerance));
     62 }
     63 
     64 template<typename VectorType,
     65          typename HyperplaneType>
     66 void check_fitHyperplane(int numPoints,
     67                          VectorType **points,
     68                          const HyperplaneType& original,
     69                          typename VectorType::Scalar tolerance)
     70 {
     71   int size = points[0]->size();
     72   HyperplaneType result(size);
     73   fitHyperplane(numPoints, points, &result);
     74   result.coeffs() *= original.coeffs().coeff(size)/result.coeffs().coeff(size);
     75   typename VectorType::Scalar error = (result.coeffs() - original.coeffs()).norm() / original.coeffs().norm();
     76   std::cout << ei_abs(error) << "  xxx   " << ei_abs(tolerance) << std::endl;
     77   VERIFY(ei_abs(error) < ei_abs(tolerance));
     78 }
     79 
     80 void test_eigen2_regression()
     81 {
     82   for(int i = 0; i < g_repeat; i++)
     83   {
     84 #ifdef EIGEN_TEST_PART_1
     85     {
     86       Vector2f points2f [1000];
     87       Vector2f *points2f_ptrs [1000];
     88       for(int i = 0; i < 1000; i++) points2f_ptrs[i] = &(points2f[i]);
     89       Vector2f coeffs2f;
     90       Hyperplane<float,2> coeffs3f;
     91       makeNoisyCohyperplanarPoints(1000, points2f_ptrs, &coeffs3f, 0.01f);
     92       coeffs2f[0] = -coeffs3f.coeffs()[0]/coeffs3f.coeffs()[1];
     93       coeffs2f[1] = -coeffs3f.coeffs()[2]/coeffs3f.coeffs()[1];
     94       CALL_SUBTEST(check_linearRegression(10, points2f_ptrs, coeffs2f, 0.05f));
     95       CALL_SUBTEST(check_linearRegression(100, points2f_ptrs, coeffs2f, 0.01f));
     96       CALL_SUBTEST(check_linearRegression(1000, points2f_ptrs, coeffs2f, 0.002f));
     97     }
     98 #endif
     99 #ifdef EIGEN_TEST_PART_2
    100     {
    101       Vector2f points2f [1000];
    102       Vector2f *points2f_ptrs [1000];
    103       for(int i = 0; i < 1000; i++) points2f_ptrs[i] = &(points2f[i]);
    104       Hyperplane<float,2> coeffs3f;
    105       makeNoisyCohyperplanarPoints(1000, points2f_ptrs, &coeffs3f, 0.01f);
    106       CALL_SUBTEST(check_fitHyperplane(10, points2f_ptrs, coeffs3f, 0.05f));
    107       CALL_SUBTEST(check_fitHyperplane(100, points2f_ptrs, coeffs3f, 0.01f));
    108       CALL_SUBTEST(check_fitHyperplane(1000, points2f_ptrs, coeffs3f, 0.002f));
    109     }
    110 #endif
    111 #ifdef EIGEN_TEST_PART_3
    112     {
    113       Vector4d points4d [1000];
    114       Vector4d *points4d_ptrs [1000];
    115       for(int i = 0; i < 1000; i++) points4d_ptrs[i] = &(points4d[i]);
    116       Hyperplane<double,4> coeffs5d;
    117       makeNoisyCohyperplanarPoints(1000, points4d_ptrs, &coeffs5d, 0.01);
    118       CALL_SUBTEST(check_fitHyperplane(10, points4d_ptrs, coeffs5d, 0.05));
    119       CALL_SUBTEST(check_fitHyperplane(100, points4d_ptrs, coeffs5d, 0.01));
    120       CALL_SUBTEST(check_fitHyperplane(1000, points4d_ptrs, coeffs5d, 0.002));
    121     }
    122 #endif
    123 #ifdef EIGEN_TEST_PART_4
    124     {
    125       VectorXcd *points11cd_ptrs[1000];
    126       for(int i = 0; i < 1000; i++) points11cd_ptrs[i] = new VectorXcd(11);
    127       Hyperplane<std::complex<double>,Dynamic> *coeffs12cd = new Hyperplane<std::complex<double>,Dynamic>(11);
    128       makeNoisyCohyperplanarPoints(1000, points11cd_ptrs, coeffs12cd, 0.01);
    129       CALL_SUBTEST(check_fitHyperplane(100, points11cd_ptrs, *coeffs12cd, 0.025));
    130       CALL_SUBTEST(check_fitHyperplane(1000, points11cd_ptrs, *coeffs12cd, 0.006));
    131       delete coeffs12cd;
    132       for(int i = 0; i < 1000; i++) delete points11cd_ptrs[i];
    133     }
    134 #endif
    135   }
    136 }
    137