| 1 | #region License Information
|
|---|
| 2 |
|
|---|
| 3 | /* HeuristicLab
|
|---|
| 4 | * Copyright (C) 2002-2014 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
|
|---|
| 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 |
|
|---|
| 22 | #endregion
|
|---|
| 23 |
|
|---|
| 24 | using System;
|
|---|
| 25 | using System.Linq;
|
|---|
| 26 | using HeuristicLab.Analysis;
|
|---|
| 27 | using HeuristicLab.Common;
|
|---|
| 28 | using HeuristicLab.Core;
|
|---|
| 29 | using HeuristicLab.Data;
|
|---|
| 30 | using HeuristicLab.Operators;
|
|---|
| 31 | using HeuristicLab.Optimization.Operators;
|
|---|
| 32 | using HeuristicLab.Parameters;
|
|---|
| 33 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
|---|
| 34 |
|
|---|
| 35 | namespace HeuristicLab.Problems.DataAnalysis.Symbolic {
|
|---|
| 36 | [StorableClass]
|
|---|
| 37 | [Item("SymbolicDataAnalysisSingleObjectivePruningAnalyzer", "An analyzer that prunes introns from trees in single objective symbolic data analysis problems.")]
|
|---|
| 38 | public abstract class SymbolicDataAnalysisSingleObjectivePruningAnalyzer : SymbolicDataAnalysisSingleObjectiveAnalyzer {
|
|---|
| 39 | #region parameter names
|
|---|
| 40 | private const string ProblemDataParameterName = "ProblemData";
|
|---|
| 41 | private const string UpdateIntervalParameterName = "UpdateInverval";
|
|---|
| 42 | private const string UpdateCounterParameterName = "UpdateCounter";
|
|---|
| 43 | private const string PopulationSliceParameterName = "PopulationSlice";
|
|---|
| 44 | private const string PruningProbabilityParameterName = "PruningProbability";
|
|---|
| 45 | private const string TotalNumberOfPrunedSubtreesParameterName = "Number of pruned subtrees";
|
|---|
| 46 | private const string TotalNumberOfPrunedTreesParameterName = "Number of pruned trees";
|
|---|
| 47 | private const string RandomParameterName = "Random";
|
|---|
| 48 | private const string PruneOnlyZeroImpactNodesParameterName = "PruneOnlyZeroImpactNodes";
|
|---|
| 49 | private const string NodeImpactThresholdParameterName = "ImpactThreshold";
|
|---|
| 50 | private const string PruningOperatorParameterName = "PruningOperator";
|
|---|
| 51 | private const string ResultsParameterName = "Results";
|
|---|
| 52 | private const string PopulationSizeParameterName = "PopulationSize";
|
|---|
| 53 | #endregion
|
|---|
| 54 |
|
|---|
| 55 | #region private members
|
|---|
| 56 | private DataReducer prunedSubtreesReducer;
|
|---|
| 57 | private DataReducer prunedTreesReducer;
|
|---|
| 58 | private DataTableValuesCollector valuesCollector;
|
|---|
| 59 | private ResultsCollector resultsCollector;
|
|---|
| 60 | #endregion
|
|---|
| 61 |
|
|---|
| 62 | #region parameter properties
|
|---|
| 63 | public IValueParameter<SymbolicDataAnalysisExpressionPruningOperator> PruningOperatorParameter {
|
|---|
| 64 | get { return (IValueParameter<SymbolicDataAnalysisExpressionPruningOperator>)Parameters[PruningOperatorParameterName]; }
|
|---|
| 65 | }
|
|---|
| 66 | public IFixedValueParameter<BoolValue> PruneOnlyZeroImpactNodesParameter {
|
|---|
| 67 | get { return (IFixedValueParameter<BoolValue>)Parameters[PruneOnlyZeroImpactNodesParameterName]; }
|
|---|
| 68 | }
|
|---|
| 69 | public IFixedValueParameter<DoubleValue> NodeImpactThresholdParameter {
|
|---|
| 70 | get { return (IFixedValueParameter<DoubleValue>)Parameters[NodeImpactThresholdParameterName]; }
|
|---|
| 71 | }
|
|---|
| 72 | public ILookupParameter<IRandom> RandomParameter {
|
|---|
| 73 | get { return (ILookupParameter<IRandom>)Parameters[RandomParameterName]; }
|
|---|
| 74 | }
|
|---|
| 75 | public IFixedValueParameter<IntValue> UpdateIntervalParameter {
|
|---|
| 76 | get { return (IFixedValueParameter<IntValue>)Parameters[UpdateIntervalParameterName]; }
|
|---|
| 77 | }
|
|---|
| 78 | public IFixedValueParameter<IntValue> UpdateCounterParameter {
|
|---|
| 79 | get { return (IFixedValueParameter<IntValue>)Parameters[UpdateCounterParameterName]; }
|
|---|
| 80 | }
|
|---|
| 81 | public IFixedValueParameter<DoubleRange> PopulationSliceParameter {
|
|---|
| 82 | get { return (IFixedValueParameter<DoubleRange>)Parameters[PopulationSliceParameterName]; }
|
|---|
| 83 | }
|
|---|
| 84 | public IFixedValueParameter<DoubleValue> PruningProbabilityParameter {
|
|---|
| 85 | get { return (IFixedValueParameter<DoubleValue>)Parameters[PruningProbabilityParameterName]; }
|
|---|
| 86 | }
|
|---|
| 87 | public ILookupParameter<IntValue> PopulationSizeParameter {
|
|---|
| 88 | get { return (ILookupParameter<IntValue>)Parameters[PopulationSizeParameterName]; }
|
|---|
| 89 | }
|
|---|
| 90 | #endregion
|
|---|
| 91 |
|
|---|
| 92 | #region properties
|
|---|
| 93 | protected SymbolicDataAnalysisExpressionPruningOperator PruningOperator { get { return PruningOperatorParameter.Value; } }
|
|---|
| 94 | protected int UpdateInterval { get { return UpdateIntervalParameter.Value.Value; } }
|
|---|
| 95 |
|
|---|
| 96 | protected int UpdateCounter {
|
|---|
| 97 | get { return UpdateCounterParameter.Value.Value; }
|
|---|
| 98 | set { UpdateCounterParameter.Value.Value = value; }
|
|---|
| 99 | }
|
|---|
| 100 |
|
|---|
| 101 | protected double PopulationSliceStart {
|
|---|
| 102 | get { return PopulationSliceParameter.Value.Start; }
|
|---|
| 103 | set { PopulationSliceParameter.Value.Start = value; }
|
|---|
| 104 | }
|
|---|
| 105 |
|
|---|
| 106 | protected double PopulationSliceEnd {
|
|---|
| 107 | get { return PopulationSliceParameter.Value.End; }
|
|---|
| 108 | set { PopulationSliceParameter.Value.End = value; }
|
|---|
| 109 | }
|
|---|
| 110 |
|
|---|
| 111 | protected double PruningProbability {
|
|---|
| 112 | get { return PruningProbabilityParameter.Value.Value; }
|
|---|
| 113 | set { PruningProbabilityParameter.Value.Value = value; }
|
|---|
| 114 | }
|
|---|
| 115 |
|
|---|
| 116 | protected bool PruneOnlyZeroImpactNodes {
|
|---|
| 117 | get { return PruneOnlyZeroImpactNodesParameter.Value.Value; }
|
|---|
| 118 | set { PruneOnlyZeroImpactNodesParameter.Value.Value = value; }
|
|---|
| 119 | }
|
|---|
| 120 | protected double NodeImpactThreshold {
|
|---|
| 121 | get { return NodeImpactThresholdParameter.Value.Value; }
|
|---|
| 122 | set { NodeImpactThresholdParameter.Value.Value = value; }
|
|---|
| 123 | }
|
|---|
| 124 | #endregion
|
|---|
| 125 |
|
|---|
| 126 | #region IStatefulItem members
|
|---|
| 127 | public override void InitializeState() {
|
|---|
| 128 | base.InitializeState();
|
|---|
| 129 | UpdateCounter = 0;
|
|---|
| 130 | }
|
|---|
| 131 | public override void ClearState() {
|
|---|
| 132 | base.ClearState();
|
|---|
| 133 | UpdateCounter = 0;
|
|---|
| 134 | }
|
|---|
| 135 | #endregion
|
|---|
| 136 |
|
|---|
| 137 | [StorableConstructor]
|
|---|
| 138 | protected SymbolicDataAnalysisSingleObjectivePruningAnalyzer(bool deserializing) : base(deserializing) { }
|
|---|
| 139 |
|
|---|
| 140 | protected SymbolicDataAnalysisSingleObjectivePruningAnalyzer(SymbolicDataAnalysisSingleObjectivePruningAnalyzer original, Cloner cloner)
|
|---|
| 141 | : base(original, cloner) {
|
|---|
| 142 | if (original.prunedSubtreesReducer != null)
|
|---|
| 143 | this.prunedSubtreesReducer = (DataReducer)original.prunedSubtreesReducer.Clone();
|
|---|
| 144 | if (original.prunedTreesReducer != null)
|
|---|
| 145 | this.prunedTreesReducer = (DataReducer)original.prunedTreesReducer.Clone();
|
|---|
| 146 | if (original.valuesCollector != null)
|
|---|
| 147 | this.valuesCollector = (DataTableValuesCollector)original.valuesCollector.Clone();
|
|---|
| 148 | if (original.resultsCollector != null)
|
|---|
| 149 | this.resultsCollector = (ResultsCollector)original.resultsCollector.Clone();
|
|---|
| 150 | }
|
|---|
| 151 |
|
|---|
| 152 | [StorableHook(HookType.AfterDeserialization)]
|
|---|
| 153 | private void AfterDeserialization() {
|
|---|
| 154 | if (!Parameters.ContainsKey(PopulationSizeParameterName)) {
|
|---|
| 155 | Parameters.Add(new LookupParameter<IntValue>(PopulationSizeParameterName, "The population of individuals."));
|
|---|
| 156 | }
|
|---|
| 157 | if (Parameters.ContainsKey(UpdateCounterParameterName)) {
|
|---|
| 158 | var fixedValueParameter = Parameters[UpdateCounterParameterName] as FixedValueParameter<IntValue>;
|
|---|
| 159 | if (fixedValueParameter == null) {
|
|---|
| 160 | var valueParameter = (ValueParameter<IntValue>)Parameters[UpdateCounterParameterName];
|
|---|
| 161 | Parameters.Remove(UpdateCounterParameterName);
|
|---|
| 162 | Parameters.Add(new FixedValueParameter<IntValue>(UpdateCounterParameterName, valueParameter.Value));
|
|---|
| 163 | }
|
|---|
| 164 | }
|
|---|
| 165 | if (Parameters.ContainsKey(UpdateIntervalParameterName)) {
|
|---|
| 166 | var fixedValueParameter = Parameters[UpdateIntervalParameterName] as FixedValueParameter<IntValue>;
|
|---|
| 167 | if (fixedValueParameter == null) {
|
|---|
| 168 | var valueParameter = (ValueParameter<IntValue>)Parameters[UpdateIntervalParameterName];
|
|---|
| 169 | Parameters.Remove(UpdateIntervalParameterName);
|
|---|
| 170 | Parameters.Add(new FixedValueParameter<IntValue>(UpdateIntervalParameterName, valueParameter.Value));
|
|---|
| 171 | }
|
|---|
| 172 | }
|
|---|
| 173 | if (Parameters.ContainsKey(PopulationSliceParameterName)) {
|
|---|
| 174 | var fixedValueParameter = Parameters[PopulationSliceParameterName] as FixedValueParameter<DoubleRange>;
|
|---|
| 175 | if (fixedValueParameter == null) {
|
|---|
| 176 | var valueParameter = (ValueParameter<DoubleRange>)Parameters[PopulationSliceParameterName];
|
|---|
| 177 | Parameters.Remove(PopulationSliceParameterName);
|
|---|
| 178 | Parameters.Add(new FixedValueParameter<DoubleRange>(PopulationSliceParameterName, valueParameter.Value));
|
|---|
| 179 | }
|
|---|
| 180 | }
|
|---|
| 181 | if (Parameters.ContainsKey(PruningProbabilityParameterName)) {
|
|---|
| 182 | var fixedValueParameter = Parameters[PruningProbabilityParameterName] as FixedValueParameter<DoubleValue>;
|
|---|
| 183 | if (fixedValueParameter == null) {
|
|---|
| 184 | var valueParameter = (ValueParameter<DoubleValue>)Parameters[PruningProbabilityParameterName];
|
|---|
| 185 | Parameters.Remove(PruningProbabilityParameterName);
|
|---|
| 186 | Parameters.Add(new FixedValueParameter<DoubleValue>(PruningProbabilityParameterName, valueParameter.Value));
|
|---|
| 187 | }
|
|---|
| 188 | }
|
|---|
| 189 | }
|
|---|
| 190 |
|
|---|
| 191 | protected SymbolicDataAnalysisSingleObjectivePruningAnalyzer() {
|
|---|
| 192 | #region add parameters
|
|---|
| 193 | Parameters.Add(new FixedValueParameter<DoubleRange>(PopulationSliceParameterName, "The slice of the population where pruning should be applied.", new DoubleRange(0.75, 1)));
|
|---|
| 194 | Parameters.Add(new FixedValueParameter<DoubleValue>(PruningProbabilityParameterName, "The probability for pruning an individual.", new DoubleValue(0.5)));
|
|---|
| 195 | Parameters.Add(new FixedValueParameter<IntValue>(UpdateIntervalParameterName, "The interval in which the tree length analysis should be applied.", new IntValue(1)));
|
|---|
| 196 | Parameters.Add(new FixedValueParameter<IntValue>(UpdateCounterParameterName, "The value which counts how many times the operator was called", new IntValue(0)));
|
|---|
| 197 | Parameters.Add(new LookupParameter<IRandom>(RandomParameterName, "The random number generator."));
|
|---|
| 198 | Parameters.Add(new LookupParameter<IDataAnalysisProblemData>(ProblemDataParameterName, "The problem data."));
|
|---|
| 199 | Parameters.Add(new FixedValueParameter<DoubleValue>(NodeImpactThresholdParameterName, "The impact threshold below which an individual should be pruned.", new DoubleValue(0.0)));
|
|---|
| 200 | Parameters.Add(new FixedValueParameter<BoolValue>(PruneOnlyZeroImpactNodesParameterName, "Switch to determine of only zero impact individuals should be pruned.", new BoolValue(false)));
|
|---|
| 201 | Parameters.Add(new LookupParameter<IntValue>(PopulationSizeParameterName, "The population of individuals."));
|
|---|
| 202 | #endregion
|
|---|
| 203 | }
|
|---|
| 204 |
|
|---|
| 205 | //
|
|---|
| 206 | /// <summary>
|
|---|
| 207 | /// Computes the closed interval bounding the portion of the population that is to be pruned.
|
|---|
| 208 | /// </summary>
|
|---|
| 209 | /// <returns>Returns an int range [start, end]</returns>
|
|---|
| 210 | private IntRange GetSliceBounds() {
|
|---|
| 211 | if (PopulationSliceStart < 0 || PopulationSliceEnd < 0) throw new ArgumentOutOfRangeException("The slice bounds cannot be negative.");
|
|---|
| 212 | if (PopulationSliceStart > 1 || PopulationSliceEnd > 1) throw new ArgumentOutOfRangeException("The slice bounds should be expressed as unit percentages.");
|
|---|
| 213 | var count = PopulationSizeParameter.ActualValue.Value;
|
|---|
| 214 | var start = (int)Math.Round(PopulationSliceStart * count);
|
|---|
| 215 | var end = (int)Math.Round(PopulationSliceEnd * count);
|
|---|
| 216 | if (end > count) end = count;
|
|---|
| 217 |
|
|---|
| 218 | if (start >= end) throw new ArgumentOutOfRangeException("Invalid PopulationSlice bounds.");
|
|---|
| 219 | return new IntRange(start, end);
|
|---|
| 220 | }
|
|---|
| 221 |
|
|---|
| 222 | private IOperation CreatePruningOperation() {
|
|---|
| 223 | var operations = new OperationCollection { Parallel = true };
|
|---|
| 224 | var range = GetSliceBounds();
|
|---|
| 225 | var qualities = Quality.Select(x => x.Value).ToArray();
|
|---|
| 226 | var indices = Enumerable.Range(0, qualities.Length).ToArray();
|
|---|
| 227 | Array.Sort(qualities, indices);
|
|---|
| 228 | if (!Maximization.Value) Array.Reverse(indices);
|
|---|
| 229 |
|
|---|
| 230 | var subscopes = ExecutionContext.Scope.SubScopes;
|
|---|
| 231 | var random = RandomParameter.ActualValue;
|
|---|
| 232 |
|
|---|
| 233 | var empty = new EmptyOperator();
|
|---|
| 234 |
|
|---|
| 235 | for (int i = 0; i < subscopes.Count; ++i) {
|
|---|
| 236 | IOperator @operator;
|
|---|
| 237 | if (range.Start <= i && i < range.End && random.NextDouble() <= PruningProbability)
|
|---|
| 238 | @operator = PruningOperator;
|
|---|
| 239 | else @operator = empty;
|
|---|
| 240 | var index = indices[i];
|
|---|
| 241 | var subscope = subscopes[index];
|
|---|
| 242 | operations.Add(ExecutionContext.CreateChildOperation(@operator, subscope));
|
|---|
| 243 | }
|
|---|
| 244 | return operations;
|
|---|
| 245 | }
|
|---|
| 246 |
|
|---|
| 247 | public override IOperation Apply() {
|
|---|
| 248 | UpdateCounter++;
|
|---|
| 249 | if (UpdateCounter != UpdateInterval) return base.Apply();
|
|---|
| 250 | UpdateCounter = 0;
|
|---|
| 251 |
|
|---|
| 252 | if (prunedSubtreesReducer == null || prunedTreesReducer == null || valuesCollector == null || resultsCollector == null) { InitializeOperators(); }
|
|---|
| 253 |
|
|---|
| 254 | var prune = CreatePruningOperation();
|
|---|
| 255 | var reducePrunedSubtrees = ExecutionContext.CreateChildOperation(prunedSubtreesReducer);
|
|---|
| 256 | var reducePrunedTrees = ExecutionContext.CreateChildOperation(prunedTreesReducer);
|
|---|
| 257 | var collectValues = ExecutionContext.CreateChildOperation(valuesCollector);
|
|---|
| 258 | var collectResults = ExecutionContext.CreateChildOperation(resultsCollector);
|
|---|
| 259 |
|
|---|
| 260 | return new OperationCollection { prune, reducePrunedSubtrees, reducePrunedTrees, collectValues, collectResults, base.Apply() };
|
|---|
| 261 | }
|
|---|
| 262 |
|
|---|
| 263 | private void InitializeOperators() {
|
|---|
| 264 | prunedSubtreesReducer = new DataReducer();
|
|---|
| 265 | prunedSubtreesReducer.ParameterToReduce.ActualName = PruningOperator.PrunedSubtreesParameter.ActualName;
|
|---|
| 266 | prunedSubtreesReducer.ReductionOperation.Value = new ReductionOperation(ReductionOperations.Sum); // sum all the pruned subtrees parameter values
|
|---|
| 267 | prunedSubtreesReducer.TargetOperation.Value = new ReductionOperation(ReductionOperations.Assign); // asign the sum to the target parameter
|
|---|
| 268 | prunedSubtreesReducer.TargetParameter.ActualName = TotalNumberOfPrunedSubtreesParameterName;
|
|---|
| 269 |
|
|---|
| 270 | prunedTreesReducer = new DataReducer();
|
|---|
| 271 | prunedTreesReducer.ParameterToReduce.ActualName = PruningOperator.PrunedTreesParameter.ActualName;
|
|---|
| 272 | prunedTreesReducer.ReductionOperation.Value = new ReductionOperation(ReductionOperations.Sum);
|
|---|
| 273 | prunedTreesReducer.TargetOperation.Value = new ReductionOperation(ReductionOperations.Assign);
|
|---|
| 274 | prunedTreesReducer.TargetParameter.ActualName = TotalNumberOfPrunedTreesParameterName;
|
|---|
| 275 |
|
|---|
| 276 | valuesCollector = new DataTableValuesCollector();
|
|---|
| 277 | valuesCollector.CollectedValues.Add(new LookupParameter<IntValue>(TotalNumberOfPrunedSubtreesParameterName));
|
|---|
| 278 | valuesCollector.CollectedValues.Add(new LookupParameter<IntValue>(TotalNumberOfPrunedTreesParameterName));
|
|---|
| 279 | valuesCollector.DataTableParameter.ActualName = "Population pruning";
|
|---|
| 280 |
|
|---|
| 281 | resultsCollector = new ResultsCollector();
|
|---|
| 282 | resultsCollector.CollectedValues.Add(new LookupParameter<DataTable>("Population pruning"));
|
|---|
| 283 | resultsCollector.ResultsParameter.ActualName = ResultsParameterName;
|
|---|
| 284 | }
|
|---|
| 285 | }
|
|---|
| 286 | }
|
|---|
