1 | #region License Information
|
---|
2 | /* HeuristicLab
|
---|
3 | * Copyright (C) 2002-2012 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
|
---|
4 | *
|
---|
5 | * This file is part of HeuristicLab.
|
---|
6 | *
|
---|
7 | * HeuristicLab is free software: you can redistribute it and/or modify
|
---|
8 | * it under the terms of the GNU General Public License as published by
|
---|
9 | * the Free Software Foundation, either version 3 of the License, or
|
---|
10 | * (at your option) any later version.
|
---|
11 | *
|
---|
12 | * HeuristicLab is distributed in the hope that it will be useful,
|
---|
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
---|
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
---|
15 | * GNU General Public License for more details.
|
---|
16 | *
|
---|
17 | * You should have received a copy of the GNU General Public License
|
---|
18 | * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.
|
---|
19 | */
|
---|
20 | #endregion
|
---|
21 |
|
---|
22 | using System;
|
---|
23 | using System.Collections.Generic;
|
---|
24 | using System.Linq;
|
---|
25 | using HeuristicLab.Common;
|
---|
26 |
|
---|
27 | namespace HeuristicLab.Problems.DataAnalysis {
|
---|
28 | public class HoeffdingsDependenceCalculator {
|
---|
29 |
|
---|
30 | public static double Calculate(IEnumerable<double> originalValues, IEnumerable<double> estimatedValues, out OnlineCalculatorError errorState) {
|
---|
31 | double d = HoeffD(originalValues, estimatedValues, out errorState);
|
---|
32 | if (errorState != OnlineCalculatorError.None) return double.NaN;
|
---|
33 | return d;
|
---|
34 | }
|
---|
35 |
|
---|
36 | /// <summary>
|
---|
37 | /// computes Hoeffding's dependence coefficient.
|
---|
38 | /// Source: hoeffd.r from R package hmisc http://cran.r-project.org/web/packages/Hmisc/index.html
|
---|
39 | /// </summary>
|
---|
40 | private static double HoeffD(IEnumerable<double> xs, IEnumerable<double> ys, out OnlineCalculatorError errorState) {
|
---|
41 | double[] rx = TiedRank(xs);
|
---|
42 | double[] ry = TiedRank(ys);
|
---|
43 | if (rx.Length != ry.Length) throw new ArgumentException("The number of elements in xs and ys does not match");
|
---|
44 | double[] rxy = TiedRank(xs, ys);
|
---|
45 |
|
---|
46 | int n = rx.Length;
|
---|
47 | double q = 0, r = 0, s = 0;
|
---|
48 | double scaling = 1.0 / (n * (n - 1));
|
---|
49 | for (int i = 0; i < n; i++) {
|
---|
50 | q += (rx[i] - 1) * (rx[i] - 2) * (ry[i] - 1) * (ry[i] - 2) * scaling;
|
---|
51 | r += (rx[i] - 2) * (ry[i] - 2) * rxy[i] * scaling;
|
---|
52 | s += rxy[i] * (rxy[i] - 1) * scaling;
|
---|
53 | }
|
---|
54 | errorState = OnlineCalculatorError.None;
|
---|
55 | // return 30.0 * (q - 2 * (n - 2) * r + (n - 2) * (n - 3) * s) / n / (n - 1) / (n - 2) / (n - 3) / (n - 4);
|
---|
56 | double t0 = q / (n - 2) / (n - 3) / (n - 4);
|
---|
57 | double t1 = 2 * r / (n - 3) / (n - 4);
|
---|
58 | double t2 = s / (n - 4);
|
---|
59 | return 30.0 * (t0 - t1 + t2);
|
---|
60 | }
|
---|
61 |
|
---|
62 | private static double[] TiedRank(IEnumerable<double> xs) {
|
---|
63 | var xsArr = xs.ToArray();
|
---|
64 | var idx = Enumerable.Range(1, xsArr.Length).ToArray();
|
---|
65 | Array.Sort(xsArr, idx);
|
---|
66 | CRank(xsArr);
|
---|
67 | Array.Sort(idx, xsArr);
|
---|
68 | return xsArr;
|
---|
69 | }
|
---|
70 |
|
---|
71 | /// <summary>
|
---|
72 | /// Calculates the joint rank with midranks for ties. Source: hoeffd.r from R package hmisc http://cran.r-project.org/web/packages/Hmisc/index.html
|
---|
73 | /// </summary>
|
---|
74 | /// <param name="xs"></param>
|
---|
75 | /// <param name="ys"></param>
|
---|
76 | /// <returns></returns>
|
---|
77 | private static double[] TiedRank(IEnumerable<double> xs, IEnumerable<double> ys) {
|
---|
78 | var xsArr = xs.ToArray();
|
---|
79 | var ysArr = ys.ToArray();
|
---|
80 | var r = new double[xsArr.Length];
|
---|
81 | int n = r.Length;
|
---|
82 | for (int i = 0; i < n; i++) {
|
---|
83 | var xi = xsArr[i];
|
---|
84 | var yi = ysArr[i];
|
---|
85 | double ri = 0.0;
|
---|
86 | for (int j = 0; j < n; j++) {
|
---|
87 | if (i != j) {
|
---|
88 | double cx;
|
---|
89 | if (xsArr[j] < xi) cx = 1.0;
|
---|
90 | else if (xsArr[j] > xi) cx = 0.0;
|
---|
91 | else cx = 0.5; // eq
|
---|
92 | double cy;
|
---|
93 | if (ysArr[j] < yi) cy = 1.0;
|
---|
94 | else if (ysArr[j] > yi) cy = 0.0;
|
---|
95 | else cy = 0.5; // eq
|
---|
96 | ri = ri + cx * cy;
|
---|
97 | }
|
---|
98 | }
|
---|
99 | r[i] = ri;
|
---|
100 | }
|
---|
101 | return r;
|
---|
102 | }
|
---|
103 |
|
---|
104 | /// <summary>
|
---|
105 | /// Calculates midranks. Source: Numerical Recipes in C.
|
---|
106 | /// </summary>
|
---|
107 | /// <param name="w">Sorted array of elements, replaces the elements by their rank, including midranking of ties</param>
|
---|
108 | /// <returns></returns>
|
---|
109 | private static void CRank(double[] w) {
|
---|
110 | int i = 0;
|
---|
111 | int n = w.Length;
|
---|
112 | while (i < n - 1) {
|
---|
113 | if (w[i + 1] > w[i]) { // w[i+1] must be larger or equal w[i] as w must be sorted
|
---|
114 | // not a tie
|
---|
115 | w[i] = i + 1;
|
---|
116 | i++;
|
---|
117 | } else {
|
---|
118 | int j;
|
---|
119 | for (j = i + 1; j < n && w[j] <= w[i]; j++) ; // how far does it go (<= effectively means == as w must be sorted)
|
---|
120 | double rank = 0.5 * (i + j - 1);
|
---|
121 | int k;
|
---|
122 | for (k = i; k < j - 1; k++) w[k] = rank; // set the rank for all tied entries
|
---|
123 | i = j;
|
---|
124 | }
|
---|
125 | }
|
---|
126 |
|
---|
127 | if (i == n - 1) w[n - 1] = n - 1; // if the last element was not tied, this is its rank
|
---|
128 | }
|
---|
129 | }
|
---|
130 | }
|
---|