[16108] | 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.Algorithms.DataAnalysis;
|
---|
| 26 | using HeuristicLab.Algorithms.EGO;
|
---|
| 27 | using HeuristicLab.Algorithms.SAPBA.Operators;
|
---|
| 28 | using HeuristicLab.Analysis;
|
---|
| 29 | using HeuristicLab.Common;
|
---|
| 30 | using HeuristicLab.Core;
|
---|
| 31 | using HeuristicLab.Data;
|
---|
| 32 | using HeuristicLab.Encodings.RealVectorEncoding;
|
---|
| 33 | using HeuristicLab.Optimization;
|
---|
| 34 | using HeuristicLab.Parameters;
|
---|
| 35 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
| 36 | using HeuristicLab.Problems.DataAnalysis;
|
---|
| 37 |
|
---|
| 38 | namespace HeuristicLab.Algorithms.SAPBA {
|
---|
| 39 | [StorableClass]
|
---|
| 40 | public class LamarckianStrategy : InfillStrategy {
|
---|
| 41 | #region Parameternames
|
---|
| 42 | public const string NoTrainingPointsParameterName = "Number of Trainingpoints";
|
---|
| 43 | public const string LocalInfillCriterionParameterName = "LocalInfillCriterion";
|
---|
| 44 | public const string OptimizationAlgorithmParameterName = "Optimization Algorithm";
|
---|
| 45 | public const string RegressionAlgorithmParameterName = "Regression Algorithm";
|
---|
| 46 | #endregion
|
---|
| 47 | #region Parameters
|
---|
| 48 | public IFixedValueParameter<IntValue> NoTrainingPointsParameter => Parameters[NoTrainingPointsParameterName] as IFixedValueParameter<IntValue>;
|
---|
| 49 | public IValueParameter<IAlgorithm> OptimizationAlgorithmParameter => Parameters[OptimizationAlgorithmParameterName] as IValueParameter<IAlgorithm>;
|
---|
| 50 | public IValueParameter<IDataAnalysisAlgorithm<IRegressionProblem>> RegressionAlgorithmParameter => Parameters[RegressionAlgorithmParameterName] as IValueParameter<IDataAnalysisAlgorithm<IRegressionProblem>>;
|
---|
| 51 | public IConstrainedValueParameter<IInfillCriterion> LocalInfillCriterionParameter => Parameters[LocalInfillCriterionParameterName] as IConstrainedValueParameter<IInfillCriterion>;
|
---|
| 52 | #endregion
|
---|
| 53 | #region Properties
|
---|
| 54 | public IntValue NoTrainingPoints => NoTrainingPointsParameter.Value;
|
---|
| 55 | public IAlgorithm OptimizationAlgorithm => OptimizationAlgorithmParameter.Value;
|
---|
| 56 | public IDataAnalysisAlgorithm<IRegressionProblem> RegressionAlgorithm => RegressionAlgorithmParameter.Value;
|
---|
| 57 | public IInfillCriterion LocalInfillCriterion => LocalInfillCriterionParameter.Value;
|
---|
| 58 | #endregion
|
---|
| 59 |
|
---|
| 60 | #region Constructors
|
---|
| 61 | [StorableConstructor]
|
---|
| 62 | protected LamarckianStrategy(bool deserializing) : base(deserializing) { }
|
---|
| 63 | [StorableHook(HookType.AfterDeserialization)]
|
---|
| 64 | private void AfterDeserialization() {
|
---|
| 65 | RegisterParameterEvents();
|
---|
| 66 | }
|
---|
| 67 | protected LamarckianStrategy(LamarckianStrategy original, Cloner cloner) : base(original, cloner) {
|
---|
| 68 | RegisterParameterEvents();
|
---|
| 69 | }
|
---|
| 70 | public LamarckianStrategy() {
|
---|
| 71 | var localCritera = new ItemSet<IInfillCriterion> { new ExpectedQuality(), new ExpectedImprovement(), new AugmentedExpectedImprovement(), new ExpectedQuantileImprovement(), new MinimalQuantileCriterium(), new PluginExpectedImprovement() };
|
---|
| 72 | var osEs = new OffspringSelectionEvolutionStrategy.OffspringSelectionEvolutionStrategy {
|
---|
| 73 | Problem = new InfillProblem(),
|
---|
| 74 | ComparisonFactor = { Value = 1.0 },
|
---|
| 75 | MaximumGenerations = { Value = 1000 },
|
---|
| 76 | MaximumEvaluatedSolutions = { Value = 100000 },
|
---|
| 77 | PlusSelection = { Value = true },
|
---|
| 78 | PopulationSize = { Value = 1 }
|
---|
| 79 | };
|
---|
| 80 | osEs.MutatorParameter.Value = osEs.MutatorParameter.ValidValues.OfType<MultiRealVectorManipulator>().First();
|
---|
| 81 | Parameters.Add(new FixedValueParameter<IntValue>(NoTrainingPointsParameterName, "The number of sample points used to create a local model", new IntValue(50)));
|
---|
| 82 | Parameters.Add(new ConstrainedValueParameter<IInfillCriterion>(LocalInfillCriterionParameterName, "The infill criterion used to cheaply evaluate points.", localCritera, localCritera.First()));
|
---|
| 83 | Parameters.Add(new ValueParameter<IAlgorithm>(OptimizationAlgorithmParameterName, "The algorithm used to solve the expected improvement subproblem", osEs));
|
---|
| 84 | Parameters.Add(new ValueParameter<IDataAnalysisAlgorithm<IRegressionProblem>>(RegressionAlgorithmParameterName, "The model used to approximate the problem", new GaussianProcessRegression { Problem = new RegressionProblem() }));
|
---|
| 85 | RegisterParameterEvents();
|
---|
| 86 | }
|
---|
| 87 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
| 88 | return new LamarckianStrategy(this, cloner);
|
---|
| 89 | }
|
---|
| 90 | #endregion
|
---|
| 91 |
|
---|
| 92 | //Short lived stores for analysis
|
---|
| 93 | private readonly List<double> LamarckValues = new List<double>();
|
---|
| 94 | private readonly List<double> SampleValues = new List<double>();
|
---|
| 95 |
|
---|
| 96 | protected override void Initialize() {
|
---|
| 97 | base.Initialize();
|
---|
| 98 | var infillProblem = OptimizationAlgorithm.Problem as InfillProblem;
|
---|
| 99 | if (infillProblem == null) throw new ArgumentException("InfillOptimizationAlgorithm does not have an InfillProblem.");
|
---|
| 100 | infillProblem.InfillCriterion = LocalInfillCriterion;
|
---|
| 101 | }
|
---|
| 102 | protected override void Analyze(Individual[] individuals, double[] qualities, ResultCollection results, ResultCollection globalResults, IRandom random) {
|
---|
| 103 | base.Analyze(individuals, qualities, results, globalResults, random);
|
---|
| 104 | const string plotName = "Lamarck Comparison";
|
---|
| 105 | const string lamarckRow = "Lamarck Values";
|
---|
| 106 | const string samplesRow = "Original Values";
|
---|
| 107 | if (!globalResults.ContainsKey(plotName))
|
---|
| 108 | globalResults.Add(new Result(plotName, new DataTable(plotName)));
|
---|
| 109 |
|
---|
| 110 | var plot = (DataTable)globalResults[plotName].Value;
|
---|
| 111 | if (!plot.Rows.ContainsKey(lamarckRow)) plot.Rows.Add(new DataRow(lamarckRow));
|
---|
| 112 | if (!plot.Rows.ContainsKey(samplesRow)) plot.Rows.Add(new DataRow(samplesRow));
|
---|
| 113 | plot.Rows[lamarckRow].Values.AddRange(LamarckValues);
|
---|
| 114 | plot.Rows[samplesRow].Values.AddRange(SampleValues);
|
---|
| 115 | LamarckValues.Clear();
|
---|
| 116 | SampleValues.Clear();
|
---|
| 117 |
|
---|
| 118 | //analyze Hypervolumes
|
---|
| 119 | const string volPlotName = "Hypervolumes Comparison";
|
---|
| 120 | const string mainRowName = "Population Volume (log)";
|
---|
| 121 | const string subspaceRowName = "Subspace Volume (log) for Lamarck Candidate ";
|
---|
| 122 | if (!globalResults.ContainsKey(volPlotName))
|
---|
| 123 | globalResults.Add(new Result(volPlotName, new DataTable(volPlotName)));
|
---|
| 124 |
|
---|
| 125 | plot = (DataTable)globalResults[volPlotName].Value;
|
---|
| 126 | if (!plot.Rows.ContainsKey(mainRowName)) plot.Rows.Add(new DataRow(mainRowName));
|
---|
| 127 | var v = Math.Log(GetStableVolume(GetBoundingBox(individuals.Select(x => x.RealVector()))));
|
---|
| 128 | plot.Rows[mainRowName].Values.Add(v);
|
---|
| 129 |
|
---|
| 130 | var indis = individuals
|
---|
| 131 | .Zip(qualities, (individual, d) => new Tuple<Individual, double>(individual, d))
|
---|
| 132 | .OrderBy(t => SapbaAlgorithm.Problem.Maximization ? -t.Item2 : t.Item2)
|
---|
| 133 | .Take(NoIndividuals.Value)
|
---|
| 134 | .Select(t => t.Item1).ToArray();
|
---|
| 135 |
|
---|
| 136 | for (var i = 0; i < indis.Length; i++) {
|
---|
| 137 | var samples = GetNearestSamples(NoTrainingPoints.Value, indis[i].RealVector());
|
---|
| 138 | var d = Math.Log(GetStableVolume(GetBoundingBox(samples.Select(x => x.Item1))));
|
---|
| 139 | if (!plot.Rows.ContainsKey(subspaceRowName + i)) plot.Rows.Add(new DataRow(subspaceRowName + i));
|
---|
| 140 | plot.Rows[subspaceRowName + i].Values.Add(d);
|
---|
| 141 | }
|
---|
| 142 |
|
---|
| 143 |
|
---|
| 144 |
|
---|
| 145 | }
|
---|
| 146 | protected override void ProcessPopulation(Individual[] individuals, double[] qualities, IRandom random) {
|
---|
| 147 | if (RegressionSolution == null) return;
|
---|
| 148 | if (Generations < NoGenerations.Value) Generations++;
|
---|
| 149 | else {
|
---|
| 150 | //Select best Individuals
|
---|
| 151 | var indis = individuals
|
---|
| 152 | .Zip(qualities, (individual, d) => new Tuple<Individual, double>(individual, d))
|
---|
| 153 | .OrderBy(t => Problem.Maximization ? -t.Item2 : t.Item2)
|
---|
| 154 | .Take(NoIndividuals.Value)
|
---|
| 155 | .Select(t => t.Item1).ToArray();
|
---|
| 156 | //Evaluate individuals
|
---|
| 157 | foreach (var individual in indis)
|
---|
| 158 | SampleValues.Add(EvaluateSample(individual.RealVector(), random).Item2);
|
---|
| 159 |
|
---|
| 160 | //Perform memetic replacement for all points
|
---|
| 161 | for (var i = 0; i < indis.Length; i++) {
|
---|
| 162 | var vector = indis[i].RealVector();
|
---|
| 163 | var altVector = OptimizeInfillProblem(vector, random);
|
---|
| 164 | LamarckValues.Add(EvaluateSample(altVector, random).Item2);
|
---|
| 165 | if (LamarckValues[i] < SampleValues[i] == Problem.Maximization) continue;
|
---|
| 166 | for (var j = 0; j < vector.Length; j++) vector[j] = altVector[j];
|
---|
| 167 | }
|
---|
| 168 |
|
---|
| 169 | BuildRegressionSolution(random);
|
---|
| 170 | Generations = 0;
|
---|
| 171 | }
|
---|
| 172 | }
|
---|
| 173 |
|
---|
| 174 | #region Events
|
---|
| 175 | private void RegisterParameterEvents() {
|
---|
| 176 | OptimizationAlgorithmParameter.ValueChanged += OnInfillAlgorithmChanged;
|
---|
| 177 | OptimizationAlgorithm.ProblemChanged += OnInfillProblemChanged;
|
---|
| 178 | LocalInfillCriterionParameter.ValueChanged += OnInfillCriterionChanged;
|
---|
| 179 | }
|
---|
| 180 | private void OnInfillCriterionChanged(object sender, EventArgs e) {
|
---|
| 181 | ((InfillProblem)OptimizationAlgorithm.Problem).InfillCriterion = LocalInfillCriterion;
|
---|
| 182 | }
|
---|
| 183 | private void OnInfillAlgorithmChanged(object sender, EventArgs e) {
|
---|
| 184 | OptimizationAlgorithm.Problem = new InfillProblem { InfillCriterion = LocalInfillCriterion };
|
---|
| 185 | OptimizationAlgorithm.ProblemChanged -= OnInfillProblemChanged; //avoid double attaching
|
---|
| 186 | OptimizationAlgorithm.ProblemChanged += OnInfillProblemChanged;
|
---|
| 187 | }
|
---|
| 188 | private void OnInfillProblemChanged(object sender, EventArgs e) {
|
---|
| 189 | OptimizationAlgorithm.ProblemChanged -= OnInfillProblemChanged;
|
---|
| 190 | OptimizationAlgorithm.Problem = new InfillProblem { InfillCriterion = LocalInfillCriterion };
|
---|
| 191 | OptimizationAlgorithm.ProblemChanged += OnInfillProblemChanged;
|
---|
| 192 | }
|
---|
| 193 | #endregion
|
---|
| 194 |
|
---|
| 195 | #region helpers
|
---|
| 196 | private RealVector OptimizeInfillProblem(RealVector point, IRandom random) {
|
---|
| 197 | var infillProblem = OptimizationAlgorithm.Problem as InfillProblem;
|
---|
| 198 | if (infillProblem == null) throw new ArgumentException("InfillOptimizationAlgorithm does not have an InfillProblem.");
|
---|
| 199 | if (infillProblem.InfillCriterion != LocalInfillCriterion) throw new ArgumentException("InfillCiriterion for Problem is not correctly set.");
|
---|
| 200 |
|
---|
| 201 | var points = Math.Min(NoTrainingPoints.Value, Samples.Count);
|
---|
| 202 | var samples = GetNearestSamples(points, point);
|
---|
| 203 | var regression = SapbaUtilities.BuildModel(samples, RegressionAlgorithm, random);
|
---|
| 204 | var box = GetBoundingBox(samples.Select(x => x.Item1));
|
---|
| 205 |
|
---|
| 206 | infillProblem.Encoding.Length = ((RealVectorEncoding)Problem.Encoding).Length;
|
---|
| 207 | infillProblem.Encoding.Bounds = box;
|
---|
| 208 | infillProblem.Encoding.SolutionCreator = new FixedRealVectorCreator(point);
|
---|
| 209 | infillProblem.Initialize(regression, Problem.Maximization);
|
---|
| 210 | var res = SapbaUtilities.SyncRunSubAlgorithm(OptimizationAlgorithm, random.Next(int.MaxValue));
|
---|
| 211 | if (!res.ContainsKey(InfillProblem.BestInfillSolutionResultName)) throw new ArgumentException("The InfillOptimizationAlgorithm did not return a best solution");
|
---|
| 212 | var v = res[InfillProblem.BestInfillSolutionResultName].Value as RealVector;
|
---|
| 213 | if (v == null) throw new ArgumentException("The InfillOptimizationAlgorithm did not return the expected result types");
|
---|
| 214 | if (!InBounds(v, box)) throw new ArgumentException("Vector not in bounds");
|
---|
| 215 | OptimizationAlgorithm.Runs.Clear();
|
---|
| 216 | return v;
|
---|
| 217 | }
|
---|
| 218 | private Tuple<RealVector, double>[] GetNearestSamples(int noSamples, RealVector point) {
|
---|
| 219 | return Samples.Select(sample => Tuple.Create(SquaredEuclidean(sample.Item1, point), sample)).OrderBy(x => x.Item1).Take(noSamples).Select(x => x.Item2).ToArray();
|
---|
| 220 | }
|
---|
| 221 | private static DoubleMatrix GetBoundingBox(IEnumerable<RealVector> samples) {
|
---|
| 222 | DoubleMatrix m = null;
|
---|
| 223 | foreach (var sample in samples)
|
---|
| 224 | if (m == null) {
|
---|
| 225 | m = new DoubleMatrix(sample.Length, 2);
|
---|
| 226 | for (var i = 0; i < sample.Length; i++) m[i, 0] = m[i, 1] = sample[i];
|
---|
| 227 | } else
|
---|
| 228 | for (var i = 0; i < sample.Length; i++) {
|
---|
| 229 | m[i, 0] = Math.Min(m[i, 0], sample[i]);
|
---|
| 230 | m[i, 1] = Math.Max(m[i, 1], sample[i]);
|
---|
| 231 | }
|
---|
| 232 | return m;
|
---|
| 233 | }
|
---|
| 234 |
|
---|
| 235 | //the volume of a bounded-box whith slightly increased dimensions (Volume can never reach 0)
|
---|
| 236 | private static double GetStableVolume(DoubleMatrix bounds) {
|
---|
| 237 | var res = 1.0;
|
---|
| 238 | for (var i = 0; i < bounds.Rows; i++) res *= bounds[i, 1] - bounds[i, 0] + 0.1;
|
---|
| 239 | return res;
|
---|
| 240 | }
|
---|
| 241 | private static bool InBounds(RealVector r, DoubleMatrix bounds) {
|
---|
| 242 | return !r.Where((t, i) => t < bounds[i, 0] || t > bounds[i, 1]).Any();
|
---|
| 243 | }
|
---|
| 244 | private static double SquaredEuclidean(RealVector a, RealVector b) {
|
---|
| 245 | return a.Select((t, i) => t - b[i]).Sum(d => d * d);
|
---|
| 246 | }
|
---|
| 247 | #endregion
|
---|
| 248 | }
|
---|
| 249 | }
|
---|