[14072] | 1 | #region License Information
|
---|
| 2 | /* HeuristicLab
|
---|
| 3 | * Copyright (C) 2002-2016 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 | using HeuristicLab.Core;
|
---|
| 27 | using HeuristicLab.Data;
|
---|
| 28 | using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
|
---|
| 29 | using HeuristicLab.Optimization;
|
---|
| 30 | using HeuristicLab.Parameters;
|
---|
| 31 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
| 32 |
|
---|
| 33 | namespace HeuristicLab.Problems.DataAnalysis.Symbolic.Regression {
|
---|
| 34 | [Item("SymbolicRegressionSingleObjectiveOSGAEvaluator", "An evaluator which tries to predict when a child will not be able to fullfil offspring selection criteria, to save evaluation time.")]
|
---|
| 35 | [StorableClass]
|
---|
| 36 | public class SymbolicRegressionSingleObjectiveOsgaEvaluator : SymbolicRegressionSingleObjectiveEvaluator {
|
---|
| 37 | private const string RelativeParentChildQualityThresholdParameterName = "RelativeParentChildQualityThreshold";
|
---|
| 38 | private const string RelativeFitnessEvaluationIntervalSizeParameterName = "RelativeFitnessEvaluationIntervalSize";
|
---|
| 39 | private const string ResultCollectionParameterName = "Results";
|
---|
[14231] | 40 | private const string AggregateStatisticsParameterName = "AggregateStatistics";
|
---|
[14072] | 41 |
|
---|
| 42 | #region parameters
|
---|
| 43 | public ILookupParameter<ResultCollection> ResultCollectionParameter {
|
---|
| 44 | get { return (ILookupParameter<ResultCollection>)Parameters[ResultCollectionParameterName]; }
|
---|
| 45 | }
|
---|
[14231] | 46 | public IValueParameter<BoolValue> AggregateStatisticsParameter {
|
---|
| 47 | get { return (IValueParameter<BoolValue>)Parameters[AggregateStatisticsParameterName]; }
|
---|
| 48 | }
|
---|
[14104] | 49 | public IValueParameter<IntMatrix> RejectedStatsParameter {
|
---|
| 50 | get { return (IValueParameter<IntMatrix>)Parameters["RejectedStats"]; }
|
---|
[14072] | 51 | }
|
---|
[14104] | 52 | public IValueParameter<IntMatrix> NotRejectedStatsParameter {
|
---|
| 53 | get { return (IValueParameter<IntMatrix>)Parameters["TotalStats"]; }
|
---|
[14072] | 54 | }
|
---|
| 55 | public IValueLookupParameter<DoubleValue> ComparisonFactorParameter {
|
---|
| 56 | get { return (ValueLookupParameter<DoubleValue>)Parameters["ComparisonFactor"]; }
|
---|
| 57 | }
|
---|
| 58 | public IFixedValueParameter<PercentValue> RelativeParentChildQualityThresholdParameter {
|
---|
| 59 | get { return (IFixedValueParameter<PercentValue>)Parameters[RelativeParentChildQualityThresholdParameterName]; }
|
---|
| 60 | }
|
---|
| 61 | public IFixedValueParameter<PercentValue> RelativeFitnessEvaluationIntervalSizeParameter {
|
---|
| 62 | get { return (IFixedValueParameter<PercentValue>)Parameters[RelativeFitnessEvaluationIntervalSizeParameterName]; }
|
---|
| 63 | }
|
---|
| 64 | public IScopeTreeLookupParameter<DoubleValue> ParentQualitiesParameter { get { return (IScopeTreeLookupParameter<DoubleValue>)Parameters["ParentQualities"]; } }
|
---|
| 65 | #endregion
|
---|
| 66 |
|
---|
| 67 | #region parameter properties
|
---|
| 68 | public double RelativeParentChildQualityThreshold {
|
---|
| 69 | get { return RelativeParentChildQualityThresholdParameter.Value.Value; }
|
---|
| 70 | set { RelativeParentChildQualityThresholdParameter.Value.Value = value; }
|
---|
| 71 | }
|
---|
| 72 |
|
---|
| 73 | public double RelativeFitnessEvaluationIntervalSize {
|
---|
| 74 | get { return RelativeFitnessEvaluationIntervalSizeParameter.Value.Value; }
|
---|
| 75 | set { RelativeFitnessEvaluationIntervalSizeParameter.Value.Value = value; }
|
---|
| 76 | }
|
---|
| 77 |
|
---|
[14104] | 78 | public IntMatrix RejectedStats {
|
---|
| 79 | get { return RejectedStatsParameter.Value; }
|
---|
| 80 | set { RejectedStatsParameter.Value = value; }
|
---|
[14072] | 81 | }
|
---|
| 82 |
|
---|
[14104] | 83 | public IntMatrix TotalStats {
|
---|
| 84 | get { return NotRejectedStatsParameter.Value; }
|
---|
| 85 | set { NotRejectedStatsParameter.Value = value; }
|
---|
[14072] | 86 | }
|
---|
| 87 | #endregion
|
---|
| 88 |
|
---|
| 89 | public override bool Maximization {
|
---|
| 90 | get { return true; }
|
---|
| 91 | }
|
---|
| 92 |
|
---|
| 93 | public SymbolicRegressionSingleObjectiveOsgaEvaluator() {
|
---|
| 94 | Parameters.Add(new ValueLookupParameter<DoubleValue>("ComparisonFactor", "Determines if the quality should be compared to the better parent (1.0), to the worse (0.0) or to any linearly interpolated value between them."));
|
---|
[14104] | 95 | Parameters.Add(new FixedValueParameter<PercentValue>(RelativeParentChildQualityThresholdParameterName, new PercentValue(0.9)));
|
---|
[14072] | 96 | Parameters.Add(new FixedValueParameter<PercentValue>(RelativeFitnessEvaluationIntervalSizeParameterName, new PercentValue(0.1)));
|
---|
| 97 | Parameters.Add(new LookupParameter<ResultCollection>(ResultCollectionParameterName));
|
---|
| 98 | Parameters.Add(new ScopeTreeLookupParameter<DoubleValue>("ParentQualities") { ActualName = "Quality" });
|
---|
[14104] | 99 | Parameters.Add(new ValueParameter<IntMatrix>("RejectedStats", new IntMatrix()));
|
---|
| 100 | Parameters.Add(new ValueParameter<IntMatrix>("TotalStats", new IntMatrix()));
|
---|
[14231] | 101 | Parameters.Add(new ValueParameter<BoolValue>(AggregateStatisticsParameterName, new BoolValue(false)));
|
---|
[14072] | 102 | }
|
---|
| 103 |
|
---|
| 104 | [StorableHook(HookType.AfterDeserialization)]
|
---|
| 105 | private void AfterDeserialization() {
|
---|
| 106 | if (!Parameters.ContainsKey(ResultCollectionParameterName))
|
---|
| 107 | Parameters.Add(new LookupParameter<ResultCollection>(ResultCollectionParameterName));
|
---|
| 108 |
|
---|
| 109 | if (!Parameters.ContainsKey("ParentQualities"))
|
---|
| 110 | Parameters.Add(new ScopeTreeLookupParameter<DoubleValue>("ParentQualities") { ActualName = "Quality" });
|
---|
[14104] | 111 |
|
---|
| 112 | if (!Parameters.ContainsKey("RejectedStats"))
|
---|
| 113 | Parameters.Add(new ValueParameter<IntMatrix>("RejectedStats", new IntMatrix()));
|
---|
| 114 |
|
---|
| 115 | if (!Parameters.ContainsKey("TotalStats"))
|
---|
| 116 | Parameters.Add(new ValueParameter<IntMatrix>("TotalStats", new IntMatrix()));
|
---|
[14072] | 117 | }
|
---|
| 118 |
|
---|
| 119 | [StorableConstructor]
|
---|
| 120 | protected SymbolicRegressionSingleObjectiveOsgaEvaluator(bool deserializing) : base(deserializing) { }
|
---|
| 121 |
|
---|
| 122 | protected SymbolicRegressionSingleObjectiveOsgaEvaluator(SymbolicRegressionSingleObjectiveOsgaEvaluator original, Cloner cloner) : base(original, cloner) { }
|
---|
| 123 |
|
---|
| 124 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
| 125 | return new SymbolicRegressionSingleObjectiveOsgaEvaluator(this, cloner);
|
---|
| 126 | }
|
---|
| 127 |
|
---|
| 128 | public override void ClearState() {
|
---|
| 129 | base.ClearState();
|
---|
[14104] | 130 | RejectedStats = new IntMatrix();
|
---|
| 131 | TotalStats = new IntMatrix();
|
---|
[14072] | 132 | }
|
---|
| 133 |
|
---|
| 134 | public override IOperation InstrumentedApply() {
|
---|
| 135 | var solution = SymbolicExpressionTreeParameter.ActualValue;
|
---|
| 136 | IEnumerable<int> rows = GenerateRowsToEvaluate();
|
---|
| 137 |
|
---|
| 138 | var interpreter = SymbolicDataAnalysisTreeInterpreterParameter.ActualValue;
|
---|
| 139 | var estimationLimits = EstimationLimitsParameter.ActualValue;
|
---|
| 140 | var problemData = ProblemDataParameter.ActualValue;
|
---|
| 141 | var applyLinearScaling = ApplyLinearScalingParameter.ActualValue.Value;
|
---|
| 142 |
|
---|
| 143 | double quality;
|
---|
| 144 | var parentQualities = ParentQualitiesParameter.ActualValue;
|
---|
| 145 |
|
---|
| 146 | // parent subscopes are not present during evaluation of the initial population
|
---|
| 147 | if (parentQualities.Length > 0) {
|
---|
| 148 | quality = Calculate(interpreter, solution, estimationLimits, problemData, rows, applyLinearScaling);
|
---|
| 149 | } else {
|
---|
| 150 | quality = Calculate(interpreter, solution, estimationLimits.Lower, estimationLimits.Upper, problemData, rows, applyLinearScaling);
|
---|
| 151 | }
|
---|
| 152 | QualityParameter.ActualValue = new DoubleValue(quality);
|
---|
| 153 |
|
---|
| 154 | return base.InstrumentedApply();
|
---|
| 155 | }
|
---|
| 156 |
|
---|
| 157 | public static double Calculate(ISymbolicDataAnalysisExpressionTreeInterpreter interpreter, ISymbolicExpressionTree solution, double lowerEstimationLimit, double upperEstimationLimit, IRegressionProblemData problemData, IEnumerable<int> rows, bool applyLinearScaling) {
|
---|
| 158 | IEnumerable<double> estimatedValues = interpreter.GetSymbolicExpressionTreeValues(solution, problemData.Dataset, rows);
|
---|
| 159 | IEnumerable<double> targetValues = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, rows);
|
---|
| 160 | OnlineCalculatorError errorState;
|
---|
| 161 |
|
---|
| 162 | double r;
|
---|
| 163 | if (applyLinearScaling) {
|
---|
| 164 | var rCalculator = new OnlinePearsonsRCalculator();
|
---|
| 165 | CalculateWithScaling(targetValues, estimatedValues, lowerEstimationLimit, upperEstimationLimit, rCalculator, problemData.Dataset.Rows);
|
---|
| 166 | errorState = rCalculator.ErrorState;
|
---|
| 167 | r = rCalculator.R;
|
---|
| 168 | } else {
|
---|
| 169 | IEnumerable<double> boundedEstimatedValues = estimatedValues.LimitToRange(lowerEstimationLimit, upperEstimationLimit);
|
---|
| 170 | r = OnlinePearsonsRCalculator.Calculate(targetValues, boundedEstimatedValues, out errorState);
|
---|
| 171 | }
|
---|
| 172 | if (errorState != OnlineCalculatorError.None) return double.NaN;
|
---|
| 173 | return r * r;
|
---|
| 174 | }
|
---|
| 175 |
|
---|
| 176 | private double Calculate(ISymbolicDataAnalysisExpressionTreeInterpreter interpreter, ISymbolicExpressionTree solution, DoubleLimit estimationLimits, IRegressionProblemData problemData, IEnumerable<int> rows, bool applyLinearScaling) {
|
---|
[14184] | 177 | var estimatedValues = interpreter.GetSymbolicExpressionTreeValues(solution, problemData.Dataset, rows).LimitToRange(estimationLimits.Lower, estimationLimits.Upper);
|
---|
[14072] | 178 | var targetValues = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, rows);
|
---|
| 179 |
|
---|
| 180 | var parentQualities = ParentQualitiesParameter.ActualValue.Select(x => x.Value);
|
---|
| 181 | var minQuality = parentQualities.Min();
|
---|
| 182 | var maxQuality = parentQualities.Max();
|
---|
| 183 | var comparisonFactor = ComparisonFactorParameter.ActualValue.Value;
|
---|
| 184 | var parentQuality = minQuality + (maxQuality - minQuality) * comparisonFactor;
|
---|
| 185 | var threshold = parentQuality * RelativeParentChildQualityThreshold;
|
---|
| 186 |
|
---|
[14184] | 187 | var pearsonRCalculator = new OnlinePearsonsRCalculator();
|
---|
| 188 | var targetValuesEnumerator = targetValues.GetEnumerator();
|
---|
| 189 | var estimatedValuesEnumerator = estimatedValues.GetEnumerator();
|
---|
[14231] | 190 | var trainingPartitionSize = problemData.TrainingPartition.Size;
|
---|
| 191 | var interval = (int)Math.Floor(trainingPartitionSize * RelativeFitnessEvaluationIntervalSize);
|
---|
[14072] | 192 |
|
---|
[14231] | 193 | var aggregateStatistics = AggregateStatisticsParameter.Value.Value;
|
---|
[14184] | 194 | var i = 0;
|
---|
[14231] | 195 | if (aggregateStatistics) {
|
---|
| 196 | var trainingEnd = problemData.TrainingPartition.End;
|
---|
| 197 | var qualityPerInterval = new List<double>();
|
---|
| 198 | while (targetValuesEnumerator.MoveNext() && estimatedValuesEnumerator.MoveNext()) {
|
---|
| 199 | pearsonRCalculator.Add(targetValuesEnumerator.Current, estimatedValuesEnumerator.Current);
|
---|
| 200 | ++i;
|
---|
| 201 | if (i % interval == 0 || i == trainingPartitionSize) {
|
---|
| 202 | var q = pearsonRCalculator.ErrorState != OnlineCalculatorError.None ? double.NaN : pearsonRCalculator.R;
|
---|
| 203 | qualityPerInterval.Add(q * q);
|
---|
| 204 | }
|
---|
[14072] | 205 | }
|
---|
[14231] | 206 | var r = pearsonRCalculator.ErrorState != OnlineCalculatorError.None ? double.NaN : pearsonRCalculator.R;
|
---|
| 207 | var actualQuality = r * r;
|
---|
| 208 |
|
---|
| 209 | bool predictedRejected = false;
|
---|
| 210 |
|
---|
| 211 | i = 0;
|
---|
| 212 | double quality = actualQuality;
|
---|
| 213 | foreach (var q in qualityPerInterval) {
|
---|
| 214 | if (double.IsNaN(q) || !(q > threshold)) {
|
---|
| 215 | predictedRejected = true;
|
---|
| 216 | quality = q;
|
---|
| 217 | break;
|
---|
| 218 | }
|
---|
| 219 | ++i;
|
---|
| 220 | }
|
---|
| 221 |
|
---|
| 222 | var actuallyRejected = !(actualQuality > parentQuality);
|
---|
| 223 |
|
---|
| 224 | if (RejectedStats.Rows == 0 || TotalStats.Rows == 0) {
|
---|
| 225 | RejectedStats = new IntMatrix(2, qualityPerInterval.Count);
|
---|
| 226 | RejectedStats.RowNames = new[] { "Predicted", "Actual" };
|
---|
| 227 | RejectedStats.ColumnNames = Enumerable.Range(1, RejectedStats.Columns).Select(x => string.Format("0-{0}", Math.Min(trainingEnd, x * interval)));
|
---|
| 228 | TotalStats = new IntMatrix(2, 2);
|
---|
| 229 | TotalStats.RowNames = new[] { "Predicted", "Actual" };
|
---|
| 230 | TotalStats.ColumnNames = new[] { "Rejected", "Not Rejected" };
|
---|
| 231 | }
|
---|
| 232 | // gather some statistics
|
---|
| 233 | if (predictedRejected) {
|
---|
| 234 | RejectedStats[0, i]++;
|
---|
| 235 | TotalStats[0, 0]++;
|
---|
| 236 | } else {
|
---|
| 237 | TotalStats[0, 1]++;
|
---|
| 238 | }
|
---|
| 239 | if (actuallyRejected) {
|
---|
| 240 | TotalStats[1, 0]++;
|
---|
| 241 | } else {
|
---|
| 242 | TotalStats[1, 1]++;
|
---|
| 243 | }
|
---|
| 244 | if (predictedRejected && actuallyRejected) {
|
---|
| 245 | RejectedStats[1, i]++;
|
---|
| 246 | }
|
---|
| 247 | return quality;
|
---|
| 248 | } else {
|
---|
| 249 | while (targetValuesEnumerator.MoveNext() && estimatedValuesEnumerator.MoveNext()) {
|
---|
| 250 | pearsonRCalculator.Add(targetValuesEnumerator.Current, estimatedValuesEnumerator.Current);
|
---|
| 251 | ++i;
|
---|
| 252 | if (i % interval == 0 || i == trainingPartitionSize) {
|
---|
| 253 | var q = pearsonRCalculator.ErrorState != OnlineCalculatorError.None ? double.NaN : pearsonRCalculator.R;
|
---|
| 254 | var quality = q * q;
|
---|
| 255 | if (!(quality > threshold))
|
---|
| 256 | return quality;
|
---|
| 257 | }
|
---|
| 258 | }
|
---|
| 259 | var r = pearsonRCalculator.ErrorState != OnlineCalculatorError.None ? double.NaN : pearsonRCalculator.R;
|
---|
| 260 | var actualQuality = r * r;
|
---|
| 261 | return actualQuality;
|
---|
[14072] | 262 | }
|
---|
| 263 | }
|
---|
| 264 |
|
---|
| 265 | public override double Evaluate(IExecutionContext context, ISymbolicExpressionTree tree, IRegressionProblemData problemData, IEnumerable<int> rows) {
|
---|
| 266 | SymbolicDataAnalysisTreeInterpreterParameter.ExecutionContext = context;
|
---|
| 267 | EstimationLimitsParameter.ExecutionContext = context;
|
---|
| 268 | ApplyLinearScalingParameter.ExecutionContext = context;
|
---|
| 269 |
|
---|
| 270 | var interpreter = SymbolicDataAnalysisTreeInterpreterParameter.ActualValue;
|
---|
| 271 | var estimationLimits = EstimationLimitsParameter.ActualValue;
|
---|
| 272 | var applyLinearScaling = ApplyLinearScalingParameter.ActualValue.Value;
|
---|
| 273 |
|
---|
| 274 | double r2 = Calculate(interpreter, tree, estimationLimits.Lower, estimationLimits.Upper, problemData, rows, applyLinearScaling);
|
---|
| 275 |
|
---|
| 276 | SymbolicDataAnalysisTreeInterpreterParameter.ExecutionContext = null;
|
---|
| 277 | EstimationLimitsParameter.ExecutionContext = null;
|
---|
| 278 | ApplyLinearScalingParameter.ExecutionContext = null;
|
---|
| 279 |
|
---|
| 280 | return r2;
|
---|
| 281 | }
|
---|
| 282 | }
|
---|
| 283 | }
|
---|