source: branches/2520_PersistenceReintegration/HeuristicLab.Algorithms.DataAnalysis/3.4/GradientBoostedTrees/LossFunctions/LogisticRegressionLoss.cs @ 16462

Last change on this file since 16462 was 16462, checked in by jkarder, 7 months ago

#2520: worked on reintegration of new persistence

  • added nuget references to HEAL.Fossil
  • added StorableType attributes to many classes
  • changed signature of StorableConstructors
  • removed some classes in old persistence
  • removed some unnecessary usings
File size: 4.0 KB
Line 
1#region License Information
2/* HeuristicLab
3 * Copyright (C) 2002-2019 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
4 * and the BEACON Center for the Study of Evolution in Action.
5 *
6 * This file is part of HeuristicLab.
7 *
8 * HeuristicLab is free software: you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation, either version 3 of the License, or
11 * (at your option) any later version.
12 *
13 * HeuristicLab is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16 * GNU General Public License for more details.
17 *
18 * You should have received a copy of the GNU General Public License
19 * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.
20 */
21#endregion
22
23using System;
24using System.Collections.Generic;
25using System.Diagnostics;
26using HeuristicLab.Common;
27using HeuristicLab.Core;
28using HEAL.Fossil;
29
30namespace HeuristicLab.Algorithms.DataAnalysis {
31  // Greedy Function Approximation: A Gradient Boosting Machine (page 9)
32  [StorableType("E91BD71E-9A1D-4352-BD68-062290F8BE9C")]
33  [Item("Logistic regression loss", "")]
34  public sealed class LogisticRegressionLoss : Item, ILossFunction {
35    public LogisticRegressionLoss() { }
36
37    public double GetLoss(IEnumerable<double> target, IEnumerable<double> pred) {
38      var targetEnum = target.GetEnumerator();
39      var predEnum = pred.GetEnumerator();
40
41      double s = 0;
42      while (targetEnum.MoveNext() & predEnum.MoveNext()) {
43        Debug.Assert(targetEnum.Current.IsAlmost(0.0) || targetEnum.Current.IsAlmost(1.0), "labels must be 0 or 1 for logistic regression loss");
44
45        var y = targetEnum.Current * 2 - 1; // y in {-1,1}
46        s += Math.Log(1 + Math.Exp(-2 * y * predEnum.Current));
47      }
48      if (targetEnum.MoveNext() | predEnum.MoveNext())
49        throw new ArgumentException("target and pred have different lengths");
50
51      return s;
52    }
53
54    public IEnumerable<double> GetLossGradient(IEnumerable<double> target, IEnumerable<double> pred) {
55      var targetEnum = target.GetEnumerator();
56      var predEnum = pred.GetEnumerator();
57
58      while (targetEnum.MoveNext() & predEnum.MoveNext()) {
59        Debug.Assert(targetEnum.Current.IsAlmost(0.0) || targetEnum.Current.IsAlmost(1.0), "labels must be 0 or 1 for logistic regression loss");
60        var y = targetEnum.Current * 2 - 1; // y in {-1,1}
61
62        yield return 2 * y / (1 + Math.Exp(2 * y * predEnum.Current));
63
64      }
65      if (targetEnum.MoveNext() | predEnum.MoveNext())
66        throw new ArgumentException("target and pred have different lengths");
67    }
68
69    // targetArr and predArr are not changed by LineSearch
70    public double LineSearch(double[] targetArr, double[] predArr, int[] idx, int startIdx, int endIdx) {
71      if (targetArr.Length != predArr.Length)
72        throw new ArgumentException("target and pred have different lengths");
73
74      // "Simple Newton-Raphson step" of eqn. 23
75      double sumY = 0.0;
76      double sumDiff = 0.0;
77      for (int i = startIdx; i <= endIdx; i++) {
78        var row = idx[i];
79        var y = targetArr[row] * 2 - 1; // y in {-1,1}
80        var pseudoResponse = 2 * y / (1 + Math.Exp(2 * y * predArr[row]));
81
82        sumY += pseudoResponse;
83        sumDiff += Math.Abs(pseudoResponse) * (2 - Math.Abs(pseudoResponse));
84      }
85      // prevent divByZero
86      sumDiff = Math.Max(1E-12, sumDiff);
87      return sumY / sumDiff;
88    }
89
90    #region item implementation
91    [StorableConstructor]
92    private LogisticRegressionLoss(StorableConstructorFlag _) : base(_) { }
93
94    private LogisticRegressionLoss(LogisticRegressionLoss original, Cloner cloner) : base(original, cloner) { }
95
96    public override IDeepCloneable Clone(Cloner cloner) {
97      return new LogisticRegressionLoss(this, cloner);
98    }
99    #endregion
100
101  }
102}
Note: See TracBrowser for help on using the repository browser.