[6577] | 1 | #region License Information
|
---|
| 2 | /* HeuristicLab
|
---|
[17181] | 3 | * Copyright (C) Heuristic and Evolutionary Algorithms Laboratory (HEAL)
|
---|
[6577] | 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;
|
---|
[15061] | 25 | using System.Threading;
|
---|
[6577] | 26 | using HeuristicLab.Common;
|
---|
| 27 | using HeuristicLab.Core;
|
---|
| 28 | using HeuristicLab.Data;
|
---|
| 29 | using HeuristicLab.Optimization;
|
---|
[10030] | 30 | using HeuristicLab.Parameters;
|
---|
[17097] | 31 | using HEAL.Attic;
|
---|
[6577] | 32 | using HeuristicLab.Problems.DataAnalysis;
|
---|
| 33 |
|
---|
| 34 | namespace HeuristicLab.Algorithms.DataAnalysis {
|
---|
| 35 | /// <summary>
|
---|
[6579] | 36 | /// Neural network classification data analysis algorithm.
|
---|
[6577] | 37 | /// </summary>
|
---|
[13297] | 38 | [Item("Neural Network Classification (NN)", "Neural network classification data analysis algorithm (wrapper for ALGLIB). Further documentation: http://www.alglib.net/dataanalysis/neuralnetworks.php")]
|
---|
[12708] | 39 | [Creatable(CreatableAttribute.Categories.DataAnalysisClassification, Priority = 130)]
|
---|
[17097] | 40 | [StorableType("8D44A55B-B162-42D6-A761-A420F4D107C4")]
|
---|
[6579] | 41 | public sealed class NeuralNetworkClassification : FixedDataAnalysisAlgorithm<IClassificationProblem> {
|
---|
[6578] | 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";
|
---|
[10030] | 47 | private const string NeuralNetworkClassificationModelResultName = "Neural network classification solution";
|
---|
[6578] | 48 |
|
---|
| 49 | #region parameter properties
|
---|
| 50 | public IFixedValueParameter<DoubleValue> DecayParameter {
|
---|
| 51 | get { return (IFixedValueParameter<DoubleValue>)Parameters[DecayParameterName]; }
|
---|
| 52 | }
|
---|
[8121] | 53 | public IConstrainedValueParameter<IntValue> HiddenLayersParameter {
|
---|
| 54 | get { return (IConstrainedValueParameter<IntValue>)Parameters[HiddenLayersParameterName]; }
|
---|
[6578] | 55 | }
|
---|
| 56 | public IFixedValueParameter<IntValue> NodesInFirstHiddenLayerParameter {
|
---|
| 57 | get { return (IFixedValueParameter<IntValue>)Parameters[NodesInFirstHiddenLayerParameterName]; }
|
---|
| 58 | }
|
---|
| 59 | public IFixedValueParameter<IntValue> NodesInSecondHiddenLayerParameter {
|
---|
| 60 | get { return (IFixedValueParameter<IntValue>)Parameters[NodesInSecondHiddenLayerParameterName]; }
|
---|
| 61 | }
|
---|
| 62 | public IFixedValueParameter<IntValue> RestartsParameter {
|
---|
| 63 | get { return (IFixedValueParameter<IntValue>)Parameters[RestartsParameterName]; }
|
---|
| 64 | }
|
---|
| 65 | #endregion
|
---|
| 66 |
|
---|
| 67 | #region properties
|
---|
| 68 | public double Decay {
|
---|
| 69 | get { return DecayParameter.Value.Value; }
|
---|
| 70 | set {
|
---|
| 71 | if (value < 0.001 || value > 100) throw new ArgumentException("The decay parameter should be set to a value between 0.001 and 100.", "Decay");
|
---|
| 72 | DecayParameter.Value.Value = value;
|
---|
| 73 | }
|
---|
| 74 | }
|
---|
| 75 | public int HiddenLayers {
|
---|
| 76 | get { return HiddenLayersParameter.Value.Value; }
|
---|
| 77 | set {
|
---|
| 78 | if (value < 0 || value > 2) throw new ArgumentException("The number of hidden layers should be set to 0, 1, or 2.", "HiddenLayers");
|
---|
| 79 | HiddenLayersParameter.Value = (from v in HiddenLayersParameter.ValidValues
|
---|
| 80 | where v.Value == value
|
---|
| 81 | select v)
|
---|
| 82 | .Single();
|
---|
| 83 | }
|
---|
| 84 | }
|
---|
| 85 | public int NodesInFirstHiddenLayer {
|
---|
| 86 | get { return NodesInFirstHiddenLayerParameter.Value.Value; }
|
---|
| 87 | set {
|
---|
| 88 | if (value < 1) throw new ArgumentException("The number of nodes in the first hidden layer must be at least one.", "NodesInFirstHiddenLayer");
|
---|
| 89 | NodesInFirstHiddenLayerParameter.Value.Value = value;
|
---|
| 90 | }
|
---|
| 91 | }
|
---|
| 92 | public int NodesInSecondHiddenLayer {
|
---|
| 93 | get { return NodesInSecondHiddenLayerParameter.Value.Value; }
|
---|
| 94 | set {
|
---|
| 95 | if (value < 1) throw new ArgumentException("The number of nodes in the first second layer must be at least one.", "NodesInSecondHiddenLayer");
|
---|
| 96 | NodesInSecondHiddenLayerParameter.Value.Value = value;
|
---|
| 97 | }
|
---|
| 98 | }
|
---|
| 99 | public int Restarts {
|
---|
| 100 | get { return RestartsParameter.Value.Value; }
|
---|
| 101 | set {
|
---|
| 102 | if (value < 0) throw new ArgumentException("The number of restarts must be positive.", "Restarts");
|
---|
| 103 | RestartsParameter.Value.Value = value;
|
---|
| 104 | }
|
---|
| 105 | }
|
---|
| 106 | #endregion
|
---|
| 107 |
|
---|
| 108 |
|
---|
[6577] | 109 | [StorableConstructor]
|
---|
[17097] | 110 | private NeuralNetworkClassification(StorableConstructorFlag _) : base(_) { }
|
---|
[6579] | 111 | private NeuralNetworkClassification(NeuralNetworkClassification original, Cloner cloner)
|
---|
[6577] | 112 | : base(original, cloner) {
|
---|
[6720] | 113 | RegisterEventHandlers();
|
---|
[6577] | 114 | }
|
---|
[6579] | 115 | public NeuralNetworkClassification()
|
---|
[6577] | 116 | : base() {
|
---|
[15788] | 117 | var validHiddenLayerValues = new ItemSet<IntValue>(new IntValue[] {
|
---|
| 118 | (IntValue)new IntValue(0).AsReadOnly(),
|
---|
| 119 | (IntValue)new IntValue(1).AsReadOnly(),
|
---|
[6720] | 120 | (IntValue)new IntValue(2).AsReadOnly() });
|
---|
[6578] | 121 | var selectedHiddenLayerValue = (from v in validHiddenLayerValues
|
---|
| 122 | where v.Value == 1
|
---|
| 123 | select v)
|
---|
| 124 | .Single();
|
---|
| 125 | 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(1)));
|
---|
| 126 | Parameters.Add(new ConstrainedValueParameter<IntValue>(HiddenLayersParameterName, "The number of hidden layers for the neural network (0, 1, or 2)", validHiddenLayerValues, selectedHiddenLayerValue));
|
---|
| 127 | Parameters.Add(new FixedValueParameter<IntValue>(NodesInFirstHiddenLayerParameterName, "The number of nodes in the first hidden layer. This value is not used if the number of hidden layers is zero.", new IntValue(10)));
|
---|
| 128 | 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(10)));
|
---|
| 129 | Parameters.Add(new FixedValueParameter<IntValue>(RestartsParameterName, "The number of restarts for learning.", new IntValue(2)));
|
---|
| 130 |
|
---|
[6720] | 131 | RestartsParameter.Hidden = true;
|
---|
| 132 | NodesInSecondHiddenLayerParameter.Hidden = true;
|
---|
| 133 |
|
---|
| 134 | RegisterEventHandlers();
|
---|
| 135 |
|
---|
[6579] | 136 | Problem = new ClassificationProblem();
|
---|
[6577] | 137 | }
|
---|
[6720] | 138 |
|
---|
| 139 | private void RegisterEventHandlers() {
|
---|
| 140 | HiddenLayersParameter.Value.ValueChanged += HiddenLayersParameterValueValueChanged;
|
---|
| 141 | HiddenLayersParameter.ValueChanged += HiddenLayersParameterValueChanged;
|
---|
| 142 | }
|
---|
| 143 |
|
---|
[6577] | 144 | [StorableHook(HookType.AfterDeserialization)]
|
---|
[6720] | 145 | private void AfterDeserialization() {
|
---|
| 146 | RegisterEventHandlers();
|
---|
| 147 | }
|
---|
[6577] | 148 |
|
---|
| 149 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
[6579] | 150 | return new NeuralNetworkClassification(this, cloner);
|
---|
[6577] | 151 | }
|
---|
[6720] | 152 | private void HiddenLayersParameterValueChanged(object source, EventArgs e) {
|
---|
| 153 | HiddenLayersParameter.Value.ValueChanged += HiddenLayersParameterValueValueChanged;
|
---|
| 154 | HiddenLayersParameterValueValueChanged(this, EventArgs.Empty);
|
---|
| 155 | }
|
---|
[6577] | 156 |
|
---|
[6720] | 157 | private void HiddenLayersParameterValueValueChanged(object source, EventArgs e) {
|
---|
| 158 | if (HiddenLayers == 0) {
|
---|
| 159 | NodesInFirstHiddenLayerParameter.Hidden = true;
|
---|
| 160 | NodesInSecondHiddenLayerParameter.Hidden = true;
|
---|
| 161 | } else if (HiddenLayers == 1) {
|
---|
| 162 | NodesInFirstHiddenLayerParameter.Hidden = false;
|
---|
| 163 | NodesInSecondHiddenLayerParameter.Hidden = true;
|
---|
| 164 | } else {
|
---|
| 165 | NodesInFirstHiddenLayerParameter.Hidden = false;
|
---|
| 166 | NodesInSecondHiddenLayerParameter.Hidden = false;
|
---|
| 167 | }
|
---|
| 168 | }
|
---|
| 169 |
|
---|
[6577] | 170 | #region neural network
|
---|
[15061] | 171 | protected override void Run(CancellationToken cancellationToken) {
|
---|
[6579] | 172 | double rmsError, avgRelError, relClassError;
|
---|
| 173 | var solution = CreateNeuralNetworkClassificationSolution(Problem.ProblemData, HiddenLayers, NodesInFirstHiddenLayer, NodesInSecondHiddenLayer, Decay, Restarts, out rmsError, out avgRelError, out relClassError);
|
---|
[10030] | 174 | Results.Add(new Result(NeuralNetworkClassificationModelResultName, "The neural network classification solution.", solution));
|
---|
| 175 | Results.Add(new Result("Root mean square error", "The root of the mean of squared errors of the neural network classification solution on the training set.", new DoubleValue(rmsError)));
|
---|
| 176 | Results.Add(new Result("Average relative error", "The average of relative errors of the neural network classification solution on the training set.", new PercentValue(avgRelError)));
|
---|
[6579] | 177 | Results.Add(new Result("Relative classification error", "The percentage of misclassified samples.", new PercentValue(relClassError)));
|
---|
[6577] | 178 | }
|
---|
| 179 |
|
---|
[6579] | 180 | public static IClassificationSolution CreateNeuralNetworkClassificationSolution(IClassificationProblemData problemData, int nLayers, int nHiddenNodes1, int nHiddenNodes2, double decay, int restarts,
|
---|
| 181 | out double rmsError, out double avgRelError, out double relClassError) {
|
---|
[12702] | 182 | var dataset = problemData.Dataset;
|
---|
[6577] | 183 | string targetVariable = problemData.TargetVariable;
|
---|
| 184 | IEnumerable<string> allowedInputVariables = problemData.AllowedInputVariables;
|
---|
[8139] | 185 | IEnumerable<int> rows = problemData.TrainingIndices;
|
---|
[15142] | 186 | double[,] inputMatrix = dataset.ToArray(allowedInputVariables.Concat(new string[] { targetVariable }), rows);
|
---|
[15788] | 187 | if (inputMatrix.ContainsNanOrInfinity())
|
---|
[6579] | 188 | throw new NotSupportedException("Neural network classification does not support NaN or infinity values in the input dataset.");
|
---|
[6577] | 189 |
|
---|
| 190 | int nRows = inputMatrix.GetLength(0);
|
---|
[6579] | 191 | int nFeatures = inputMatrix.GetLength(1) - 1;
|
---|
[6740] | 192 | double[] classValues = dataset.GetDoubleValues(targetVariable).Distinct().OrderBy(x => x).ToArray();
|
---|
[6579] | 193 | int nClasses = classValues.Count();
|
---|
| 194 | // map original class values to values [0..nClasses-1]
|
---|
[8139] | 195 | Dictionary<double, double> classIndices = new Dictionary<double, double>();
|
---|
[6579] | 196 | for (int i = 0; i < nClasses; i++) {
|
---|
[8139] | 197 | classIndices[classValues[i]] = i;
|
---|
[6579] | 198 | }
|
---|
| 199 | for (int row = 0; row < nRows; row++) {
|
---|
[8139] | 200 | inputMatrix[row, nFeatures] = classIndices[inputMatrix[row, nFeatures]];
|
---|
[6579] | 201 | }
|
---|
[6577] | 202 |
|
---|
[6580] | 203 | alglib.multilayerperceptron multiLayerPerceptron = null;
|
---|
| 204 | if (nLayers == 0) {
|
---|
| 205 | alglib.mlpcreatec0(allowedInputVariables.Count(), nClasses, out multiLayerPerceptron);
|
---|
| 206 | } else if (nLayers == 1) {
|
---|
| 207 | alglib.mlpcreatec1(allowedInputVariables.Count(), nHiddenNodes1, nClasses, out multiLayerPerceptron);
|
---|
| 208 | } else if (nLayers == 2) {
|
---|
| 209 | alglib.mlpcreatec2(allowedInputVariables.Count(), nHiddenNodes1, nHiddenNodes2, nClasses, out multiLayerPerceptron);
|
---|
| 210 | } else throw new ArgumentException("Number of layers must be zero, one, or two.", "nLayers");
|
---|
| 211 | alglib.mlpreport rep;
|
---|
| 212 |
|
---|
[6577] | 213 | int info;
|
---|
| 214 | // using mlptrainlm instead of mlptraines or mlptrainbfgs because only one parameter is necessary
|
---|
| 215 | alglib.mlptrainlm(multiLayerPerceptron, inputMatrix, nRows, decay, restarts, out info, out rep);
|
---|
[6579] | 216 | if (info != 2) throw new ArgumentException("Error in calculation of neural network classification solution");
|
---|
[6577] | 217 |
|
---|
| 218 | rmsError = alglib.mlprmserror(multiLayerPerceptron, inputMatrix, nRows);
|
---|
[6578] | 219 | avgRelError = alglib.mlpavgrelerror(multiLayerPerceptron, inputMatrix, nRows);
|
---|
[6579] | 220 | relClassError = alglib.mlpclserror(multiLayerPerceptron, inputMatrix, nRows) / (double)nRows;
|
---|
[6577] | 221 |
|
---|
[6649] | 222 | var problemDataClone = (IClassificationProblemData)problemData.Clone();
|
---|
[14027] | 223 | return new NeuralNetworkClassificationSolution(new NeuralNetworkModel(multiLayerPerceptron, targetVariable, allowedInputVariables, problemDataClone.ClassValues.ToArray()), problemDataClone);
|
---|
[6577] | 224 | }
|
---|
| 225 | #endregion
|
---|
| 226 | }
|
---|
| 227 | }
|
---|