Free cookie consent management tool by TermsFeed Policy Generator

source: stable/HeuristicLab.Algorithms.DataAnalysis/3.4/NeuralNetwork/NeuralNetworkEnsembleClassification.cs @ 12702

Last change on this file since 12702 was 12702, checked in by mkommend, 9 years ago

#2276: Merged changes to stable.

File size: 12.1 KB
Line 
1#region License Information
2/* HeuristicLab
3 * Copyright (C) 2002-2015 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
22using System;
23using System.Collections.Generic;
24using System.Linq;
25using HeuristicLab.Common;
26using HeuristicLab.Core;
27using HeuristicLab.Data;
28using HeuristicLab.Optimization;
29using HeuristicLab.Parameters;
30using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
31using HeuristicLab.Problems.DataAnalysis;
32
33namespace HeuristicLab.Algorithms.DataAnalysis {
34  /// <summary>
35  /// Neural network ensemble classification data analysis algorithm.
36  /// </summary>
37  [Item("Neural Network Ensemble Classification", "Neural network ensemble classification data analysis algorithm (wrapper for ALGLIB). Further documentation: http://www.alglib.net/dataanalysis/mlpensembles.php")]
38  [Creatable("Data Analysis")]
39  [StorableClass]
40  public sealed class NeuralNetworkEnsembleClassification : FixedDataAnalysisAlgorithm<IClassificationProblem> {
41    private const string EnsembleSizeParameterName = "EnsembleSize";
42    private const string DecayParameterName = "Decay";
43    private const string HiddenLayersParameterName = "HiddenLayers";
44    private const string NodesInFirstHiddenLayerParameterName = "NodesInFirstHiddenLayer";
45    private const string NodesInSecondHiddenLayerParameterName = "NodesInSecondHiddenLayer";
46    private const string RestartsParameterName = "Restarts";
47    private const string NeuralNetworkEnsembleClassificationModelResultName = "Neural network ensemble classification solution";
48
49    #region parameter properties
50    public IFixedValueParameter<IntValue> EnsembleSizeParameter {
51      get { return (IFixedValueParameter<IntValue>)Parameters[EnsembleSizeParameterName]; }
52    }
53    public IFixedValueParameter<DoubleValue> DecayParameter {
54      get { return (IFixedValueParameter<DoubleValue>)Parameters[DecayParameterName]; }
55    }
56    public IConstrainedValueParameter<IntValue> HiddenLayersParameter {
57      get { return (IConstrainedValueParameter<IntValue>)Parameters[HiddenLayersParameterName]; }
58    }
59    public IFixedValueParameter<IntValue> NodesInFirstHiddenLayerParameter {
60      get { return (IFixedValueParameter<IntValue>)Parameters[NodesInFirstHiddenLayerParameterName]; }
61    }
62    public IFixedValueParameter<IntValue> NodesInSecondHiddenLayerParameter {
63      get { return (IFixedValueParameter<IntValue>)Parameters[NodesInSecondHiddenLayerParameterName]; }
64    }
65    public IFixedValueParameter<IntValue> RestartsParameter {
66      get { return (IFixedValueParameter<IntValue>)Parameters[RestartsParameterName]; }
67    }
68    #endregion
69
70    #region properties
71    public int EnsembleSize {
72      get { return EnsembleSizeParameter.Value.Value; }
73      set {
74        if (value < 1) throw new ArgumentException("The number of models in the ensemble must be positive and at least one.", "EnsembleSize");
75        EnsembleSizeParameter.Value.Value = value;
76      }
77    }
78    public double Decay {
79      get { return DecayParameter.Value.Value; }
80      set {
81        if (value < 0.001 || value > 100) throw new ArgumentException("The decay parameter should be set to a value between 0.001 and 100.", "Decay");
82        DecayParameter.Value.Value = value;
83      }
84    }
85    public int HiddenLayers {
86      get { return HiddenLayersParameter.Value.Value; }
87      set {
88        if (value < 0 || value > 2) throw new ArgumentException("The number of hidden layers should be set to 0, 1, or 2.", "HiddenLayers");
89        HiddenLayersParameter.Value = (from v in HiddenLayersParameter.ValidValues
90                                       where v.Value == value
91                                       select v)
92                                      .Single();
93      }
94    }
95    public int NodesInFirstHiddenLayer {
96      get { return NodesInFirstHiddenLayerParameter.Value.Value; }
97      set {
98        if (value < 1) throw new ArgumentException("The number of nodes in the first hidden layer must be at least one.", "NodesInFirstHiddenLayer");
99        NodesInFirstHiddenLayerParameter.Value.Value = value;
100      }
101    }
102    public int NodesInSecondHiddenLayer {
103      get { return NodesInSecondHiddenLayerParameter.Value.Value; }
104      set {
105        if (value < 1) throw new ArgumentException("The number of nodes in the first second layer must be at least one.", "NodesInSecondHiddenLayer");
106        NodesInSecondHiddenLayerParameter.Value.Value = value;
107      }
108    }
109    public int Restarts {
110      get { return RestartsParameter.Value.Value; }
111      set {
112        if (value < 0) throw new ArgumentException("The number of restarts must be positive.", "Restarts");
113        RestartsParameter.Value.Value = value;
114      }
115    }
116    #endregion
117
118
119    [StorableConstructor]
120    private NeuralNetworkEnsembleClassification(bool deserializing) : base(deserializing) { }
121    private NeuralNetworkEnsembleClassification(NeuralNetworkEnsembleClassification original, Cloner cloner)
122      : base(original, cloner) {
123    }
124    public NeuralNetworkEnsembleClassification()
125      : base() {
126        var validHiddenLayerValues = new ItemSet<IntValue>(new IntValue[] {
127        (IntValue)new IntValue(0).AsReadOnly(),
128        (IntValue)new IntValue(1).AsReadOnly(),
129        (IntValue)new IntValue(2).AsReadOnly() });
130      var selectedHiddenLayerValue = (from v in validHiddenLayerValues
131                                      where v.Value == 1
132                                      select v)
133                                     .Single();
134      Parameters.Add(new FixedValueParameter<IntValue>(EnsembleSizeParameterName, "The number of simple neural network models in the ensemble. A good value is 10.", new IntValue(10)));
135      Parameters.Add(new FixedValueParameter<DoubleValue>(DecayParameterName, "The decay parameter for the training phase of the neural network. This parameter determines the strengh of regularization and should be set to a value between 0.001 (weak regularization) to 100 (very strong regularization). The correct value should be determined via cross-validation.", new DoubleValue(0.001)));
136      Parameters.Add(new ConstrainedValueParameter<IntValue>(HiddenLayersParameterName, "The number of hidden layers for the neural network (0, 1, or 2)", validHiddenLayerValues, selectedHiddenLayerValue));
137      Parameters.Add(new FixedValueParameter<IntValue>(NodesInFirstHiddenLayerParameterName, "The number of nodes in the first hidden layer. The value should be rather large (30-100 nodes) in order to make the network highly flexible and run into the early stopping criterion). This value is not used if the number of hidden layers is zero.", new IntValue(100)));
138      Parameters.Add(new FixedValueParameter<IntValue>(NodesInSecondHiddenLayerParameterName, "The number of nodes in the second hidden layer. This value is not used if the number of hidden layers is zero or one.", new IntValue(100)));
139      Parameters.Add(new FixedValueParameter<IntValue>(RestartsParameterName, "The number of restarts for learning.", new IntValue(2)));
140
141      HiddenLayersParameter.Hidden = true;
142      NodesInFirstHiddenLayerParameter.Hidden = true;
143      NodesInSecondHiddenLayerParameter.Hidden = true;
144      RestartsParameter.Hidden = true;
145
146      Problem = new ClassificationProblem();
147    }
148    [StorableHook(HookType.AfterDeserialization)]
149    private void AfterDeserialization() { }
150
151    public override IDeepCloneable Clone(Cloner cloner) {
152      return new NeuralNetworkEnsembleClassification(this, cloner);
153    }
154
155    #region neural network ensemble
156    protected override void Run() {
157      double rmsError, avgRelError, relClassError;
158      var solution = CreateNeuralNetworkEnsembleClassificationSolution(Problem.ProblemData, EnsembleSize, HiddenLayers, NodesInFirstHiddenLayer, NodesInSecondHiddenLayer, Decay, Restarts, out rmsError, out avgRelError, out relClassError);
159      Results.Add(new Result(NeuralNetworkEnsembleClassificationModelResultName, "The neural network ensemble classification solution.", solution));
160      Results.Add(new Result("Root mean square error", "The root of the mean of squared errors of the neural network ensemble classification solution on the training set.", new DoubleValue(rmsError)));
161      Results.Add(new Result("Average relative error", "The average of relative errors of the neural network ensemble classification solution on the training set.", new PercentValue(avgRelError)));
162      Results.Add(new Result("Relative classification error", "The percentage of misclassified samples.", new PercentValue(relClassError)));
163    }
164
165    public static IClassificationSolution CreateNeuralNetworkEnsembleClassificationSolution(IClassificationProblemData problemData, int ensembleSize, int nLayers, int nHiddenNodes1, int nHiddenNodes2, double decay, int restarts,
166      out double rmsError, out double avgRelError, out double relClassError) {
167      var dataset = problemData.Dataset;
168      string targetVariable = problemData.TargetVariable;
169      IEnumerable<string> allowedInputVariables = problemData.AllowedInputVariables;
170      IEnumerable<int> rows = problemData.TrainingIndices;
171      double[,] inputMatrix = AlglibUtil.PrepareInputMatrix(dataset, allowedInputVariables.Concat(new string[] { targetVariable }), rows);
172      if (inputMatrix.Cast<double>().Any(x => double.IsNaN(x) || double.IsInfinity(x)))
173        throw new NotSupportedException("Neural network ensemble classification does not support NaN or infinity values in the input dataset.");
174
175      int nRows = inputMatrix.GetLength(0);
176      int nFeatures = inputMatrix.GetLength(1) - 1;
177      double[] classValues = dataset.GetDoubleValues(targetVariable).Distinct().OrderBy(x => x).ToArray();
178      int nClasses = classValues.Count();
179      // map original class values to values [0..nClasses-1]
180      Dictionary<double, double> classIndices = new Dictionary<double, double>();
181      for (int i = 0; i < nClasses; i++) {
182        classIndices[classValues[i]] = i;
183      }
184      for (int row = 0; row < nRows; row++) {
185        inputMatrix[row, nFeatures] = classIndices[inputMatrix[row, nFeatures]];
186      }
187
188      alglib.mlpensemble mlpEnsemble = null;
189      if (nLayers == 0) {
190        alglib.mlpecreatec0(allowedInputVariables.Count(), nClasses, ensembleSize, out mlpEnsemble);
191      } else if (nLayers == 1) {
192        alglib.mlpecreatec1(allowedInputVariables.Count(), nHiddenNodes1, nClasses, ensembleSize, out mlpEnsemble);
193      } else if (nLayers == 2) {
194        alglib.mlpecreatec2(allowedInputVariables.Count(), nHiddenNodes1, nHiddenNodes2, nClasses, ensembleSize, out mlpEnsemble);
195      } else throw new ArgumentException("Number of layers must be zero, one, or two.", "nLayers");
196      alglib.mlpreport rep;
197
198      int info;
199      alglib.mlpetraines(mlpEnsemble, inputMatrix, nRows, decay, restarts, out info, out rep);
200      if (info != 6) throw new ArgumentException("Error in calculation of neural network ensemble classification solution");
201
202      rmsError = alglib.mlpermserror(mlpEnsemble, inputMatrix, nRows);
203      avgRelError = alglib.mlpeavgrelerror(mlpEnsemble, inputMatrix, nRows);
204      relClassError = alglib.mlperelclserror(mlpEnsemble, inputMatrix, nRows);
205      var problemDataClone = (IClassificationProblemData)problemData.Clone();
206      return new NeuralNetworkEnsembleClassificationSolution(problemDataClone, new NeuralNetworkEnsembleModel(mlpEnsemble, targetVariable, allowedInputVariables, problemDataClone.ClassValues.ToArray()));
207    }
208    #endregion
209  }
210}
Note: See TracBrowser for help on using the repository browser.