[14420] | 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 System.Threading;
|
---|
[14450] | 26 | using HeuristicLab.Algorithms.MemPR.Interfaces;
|
---|
[14420] | 27 | using HeuristicLab.Common;
|
---|
| 28 | using HeuristicLab.Core;
|
---|
[14551] | 29 | using HeuristicLab.Data;
|
---|
[14420] | 30 | using HeuristicLab.Encodings.BinaryVectorEncoding;
|
---|
| 31 | using HeuristicLab.Optimization;
|
---|
| 32 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
| 33 | using HeuristicLab.PluginInfrastructure;
|
---|
| 34 | using HeuristicLab.Random;
|
---|
| 35 |
|
---|
| 36 | namespace HeuristicLab.Algorithms.MemPR.Binary {
|
---|
| 37 | [Item("MemPR (binary)", "MemPR implementation for binary vectors.")]
|
---|
| 38 | [StorableClass]
|
---|
| 39 | [Creatable(CreatableAttribute.Categories.PopulationBasedAlgorithms, Priority = 999)]
|
---|
[14552] | 40 | public class BinaryMemPR : MemPRAlgorithm<ISingleObjectiveHeuristicOptimizationProblem, BinaryVector, BinaryMemPRPopulationContext, BinaryMemPRSolutionContext> {
|
---|
[14420] | 41 | [StorableConstructor]
|
---|
| 42 | protected BinaryMemPR(bool deserializing) : base(deserializing) { }
|
---|
| 43 | protected BinaryMemPR(BinaryMemPR original, Cloner cloner) : base(original, cloner) { }
|
---|
| 44 | public BinaryMemPR() {
|
---|
[14450] | 45 | foreach (var trainer in ApplicationManager.Manager.GetInstances<ISolutionModelTrainer<BinaryMemPRPopulationContext>>())
|
---|
[14420] | 46 | SolutionModelTrainerParameter.ValidValues.Add(trainer);
|
---|
[14563] | 47 |
|
---|
| 48 | if (SolutionModelTrainerParameter.ValidValues.Count > 0) {
|
---|
| 49 | var unbiased = SolutionModelTrainerParameter.ValidValues.FirstOrDefault(x => !x.Bias);
|
---|
| 50 | if (unbiased != null) SolutionModelTrainerParameter.Value = unbiased;
|
---|
| 51 | }
|
---|
| 52 |
|
---|
[14450] | 53 | foreach (var localSearch in ApplicationManager.Manager.GetInstances<ILocalSearch<BinaryMemPRSolutionContext>>()) {
|
---|
| 54 | LocalSearchParameter.ValidValues.Add(localSearch);
|
---|
[14420] | 55 | }
|
---|
| 56 | }
|
---|
| 57 |
|
---|
| 58 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
| 59 | return new BinaryMemPR(this, cloner);
|
---|
| 60 | }
|
---|
| 61 |
|
---|
[14550] | 62 | protected override bool Eq(BinaryVector a, BinaryVector b) {
|
---|
| 63 | var len = a.Length;
|
---|
[14420] | 64 | for (var i = 0; i < len; i++)
|
---|
[14550] | 65 | if (a[i] != b[i]) return false;
|
---|
[14420] | 66 | return true;
|
---|
| 67 | }
|
---|
| 68 |
|
---|
| 69 | protected override double Dist(ISingleObjectiveSolutionScope<BinaryVector> a, ISingleObjectiveSolutionScope<BinaryVector> b) {
|
---|
[14496] | 70 | return 1.0 - HammingSimilarityCalculator.CalculateSimilarity(a.Solution, b.Solution);
|
---|
[14420] | 71 | }
|
---|
| 72 |
|
---|
[14450] | 73 | protected override ISolutionSubspace<BinaryVector> CalculateSubspace(IEnumerable<BinaryVector> solutions, bool inverse = false) {
|
---|
[14420] | 74 | var pop = solutions.ToList();
|
---|
| 75 | var N = pop[0].Length;
|
---|
| 76 | var subspace = new bool[N];
|
---|
| 77 | for (var i = 0; i < N; i++) {
|
---|
| 78 | var val = pop[0][i];
|
---|
| 79 | if (inverse) subspace[i] = true;
|
---|
| 80 | for (var p = 1; p < pop.Count; p++) {
|
---|
| 81 | if (pop[p][i] != val) subspace[i] = !inverse;
|
---|
| 82 | }
|
---|
| 83 | }
|
---|
| 84 | return new BinarySolutionSubspace(subspace);
|
---|
| 85 | }
|
---|
| 86 |
|
---|
[14544] | 87 | protected override void AdaptiveWalk(ISingleObjectiveSolutionScope<BinaryVector> scope, int maxEvals, CancellationToken token, ISolutionSubspace<BinaryVector> subspace = null) {
|
---|
[14453] | 88 | var evaluations = 0;
|
---|
[14420] | 89 | var subset = subspace != null ? ((BinarySolutionSubspace)subspace).Subspace : null;
|
---|
[14453] | 90 | if (double.IsNaN(scope.Fitness)) {
|
---|
[14552] | 91 | Context.Evaluate(scope, token);
|
---|
[14453] | 92 | evaluations++;
|
---|
| 93 | }
|
---|
[14420] | 94 | SingleObjectiveSolutionScope<BinaryVector> bestOfTheWalk = null;
|
---|
| 95 | var currentScope = (SingleObjectiveSolutionScope<BinaryVector>)scope.Clone();
|
---|
| 96 | var current = currentScope.Solution;
|
---|
| 97 | var N = current.Length;
|
---|
[14544] | 98 |
|
---|
[14420] | 99 | var subN = subset != null ? subset.Count(x => x) : N;
|
---|
[14544] | 100 | if (subN == 0) return;
|
---|
[14420] | 101 | var order = Enumerable.Range(0, N).Where(x => subset == null || subset[x]).Shuffle(Context.Random).ToArray();
|
---|
| 102 |
|
---|
[14552] | 103 | var bound = Context.Maximization ? Context.Population.Max(x => x.Fitness) : Context.Population.Min(x => x.Fitness);
|
---|
[14550] | 104 | var range = Math.Abs(bound - Context.LocalOptimaLevel);
|
---|
| 105 | if (range.IsAlmost(0)) range = Math.Abs(bound * 0.05);
|
---|
| 106 | if (range.IsAlmost(0)) { // because bound = localoptimalevel = 0
|
---|
[14544] | 107 | Context.IncrementEvaluatedSolutions(evaluations);
|
---|
| 108 | return;
|
---|
| 109 | }
|
---|
[14550] | 110 |
|
---|
| 111 | var temp = -range / Math.Log(1.0 / maxEvals);
|
---|
| 112 | var endtemp = -range / Math.Log(0.1 / maxEvals);
|
---|
| 113 | var annealFactor = Math.Pow(endtemp / temp, 1.0 / maxEvals);
|
---|
[14456] | 114 | for (var iter = 0; iter < int.MaxValue; iter++) {
|
---|
[14544] | 115 | var moved = false;
|
---|
[14420] | 116 |
|
---|
| 117 | for (var i = 0; i < subN; i++) {
|
---|
| 118 | var idx = order[i];
|
---|
| 119 | var before = currentScope.Fitness;
|
---|
| 120 | current[idx] = !current[idx];
|
---|
[14552] | 121 | Context.Evaluate(currentScope, token);
|
---|
[14453] | 122 | evaluations++;
|
---|
[14420] | 123 | var after = currentScope.Fitness;
|
---|
| 124 |
|
---|
[14544] | 125 | if (Context.IsBetter(after, before) && (bestOfTheWalk == null || Context.IsBetter(after, bestOfTheWalk.Fitness))) {
|
---|
[14420] | 126 | bestOfTheWalk = (SingleObjectiveSolutionScope<BinaryVector>)currentScope.Clone();
|
---|
[14550] | 127 | if (Context.IsBetter(bestOfTheWalk, scope)) {
|
---|
| 128 | moved = false;
|
---|
| 129 | break;
|
---|
| 130 | }
|
---|
[14420] | 131 | }
|
---|
[14552] | 132 | var diff = Context.Maximization ? after - before : before - after;
|
---|
[14544] | 133 | if (diff > 0) moved = true;
|
---|
| 134 | else {
|
---|
[14550] | 135 | var prob = Math.Exp(diff / temp);
|
---|
[14544] | 136 | if (Context.Random.NextDouble() >= prob) {
|
---|
| 137 | // the move is not good enough -> undo the move
|
---|
| 138 | current[idx] = !current[idx];
|
---|
| 139 | currentScope.Fitness = before;
|
---|
[14420] | 140 | }
|
---|
| 141 | }
|
---|
[14550] | 142 | temp *= annealFactor;
|
---|
[14456] | 143 | if (evaluations >= maxEvals) break;
|
---|
[14420] | 144 | }
|
---|
[14544] | 145 | if (!moved) break;
|
---|
[14456] | 146 | if (evaluations >= maxEvals) break;
|
---|
[14420] | 147 | }
|
---|
| 148 |
|
---|
[14453] | 149 | Context.IncrementEvaluatedSolutions(evaluations);
|
---|
[14420] | 150 | scope.Adopt(bestOfTheWalk ?? currentScope);
|
---|
| 151 | }
|
---|
| 152 |
|
---|
[14544] | 153 | protected override ISingleObjectiveSolutionScope<BinaryVector> Breed(ISingleObjectiveSolutionScope<BinaryVector> p1, ISingleObjectiveSolutionScope<BinaryVector> p2, CancellationToken token) {
|
---|
| 154 | var evaluations = 0;
|
---|
| 155 | var N = p1.Solution.Length;
|
---|
| 156 |
|
---|
[14695] | 157 | var probe = Context.ToScope(null);
|
---|
[14550] | 158 |
|
---|
| 159 | var cache = new HashSet<BinaryVector>(new BinaryVectorEqualityComparer());
|
---|
| 160 | cache.Add(p1.Solution);
|
---|
| 161 | cache.Add(p2.Solution);
|
---|
| 162 |
|
---|
[14563] | 163 | var cacheHits = new Dictionary<int, int>() { { 0, 0 }, { 1, 0 }, { 2, 0 } };
|
---|
[14550] | 164 | ISingleObjectiveSolutionScope<BinaryVector> offspring = null;
|
---|
[14563] | 165 |
|
---|
[14551] | 166 | while (evaluations < N) {
|
---|
| 167 | BinaryVector c = null;
|
---|
[14563] | 168 | var xochoice = cacheHits.SampleRandom(Context.Random).Key;
|
---|
[14557] | 169 | switch (xochoice) {
|
---|
| 170 | case 0: c = NPointCrossover.Apply(Context.Random, p1.Solution, p2.Solution, new IntValue(1)); break;
|
---|
| 171 | case 1: c = NPointCrossover.Apply(Context.Random, p1.Solution, p2.Solution, new IntValue(2)); break;
|
---|
| 172 | case 2: c = UniformCrossover.Apply(Context.Random, p1.Solution, p2.Solution); break;
|
---|
| 173 | }
|
---|
[14551] | 174 | if (cache.Contains(c)) {
|
---|
[14563] | 175 | cacheHits[xochoice]++;
|
---|
| 176 | if (cacheHits[xochoice] > 10) {
|
---|
| 177 | cacheHits.Remove(xochoice);
|
---|
| 178 | if (cacheHits.Count == 0) break;
|
---|
| 179 | }
|
---|
[14550] | 180 | continue;
|
---|
[14551] | 181 | }
|
---|
[14556] | 182 | probe.Solution = c;
|
---|
[14552] | 183 | Context.Evaluate(probe, token);
|
---|
[14544] | 184 | evaluations++;
|
---|
[14551] | 185 | cache.Add(c);
|
---|
[14550] | 186 | if (offspring == null || Context.IsBetter(probe, offspring)) {
|
---|
[14551] | 187 | offspring = probe;
|
---|
| 188 | if (Context.IsBetter(offspring, p1) && Context.IsBetter(offspring, p2))
|
---|
| 189 | break;
|
---|
[14420] | 190 | }
|
---|
| 191 | }
|
---|
[14544] | 192 | Context.IncrementEvaluatedSolutions(evaluations);
|
---|
[14550] | 193 | return offspring ?? probe;
|
---|
[14420] | 194 | }
|
---|
| 195 |
|
---|
[14544] | 196 | protected override ISingleObjectiveSolutionScope<BinaryVector> Link(ISingleObjectiveSolutionScope<BinaryVector> a, ISingleObjectiveSolutionScope<BinaryVector> b, CancellationToken token, bool delink = false) {
|
---|
| 197 | var evaluations = 0;
|
---|
| 198 | var childScope = (ISingleObjectiveSolutionScope<BinaryVector>)a.Clone();
|
---|
[14420] | 199 | var child = childScope.Solution;
|
---|
| 200 | ISingleObjectiveSolutionScope<BinaryVector> best = null;
|
---|
[14544] | 201 | var cF = a.Fitness;
|
---|
[14420] | 202 | var bF = double.NaN;
|
---|
[14544] | 203 | var order = Enumerable.Range(0, child.Length)
|
---|
| 204 | .Where(x => !delink && child[x] != b.Solution[x] || delink && child[x] == b.Solution[x])
|
---|
| 205 | .Shuffle(Context.Random).ToList();
|
---|
| 206 | if (order.Count == 0) return childScope;
|
---|
| 207 |
|
---|
[14420] | 208 | while (true) {
|
---|
| 209 | var bestS = double.NaN;
|
---|
[14544] | 210 | var bestI = -1;
|
---|
| 211 | for (var i = 0; i < order.Count; i++) {
|
---|
[14420] | 212 | var idx = order[i];
|
---|
| 213 | child[idx] = !child[idx]; // move
|
---|
[14552] | 214 | Context.Evaluate(childScope, token);
|
---|
[14453] | 215 | evaluations++;
|
---|
[14420] | 216 | var s = childScope.Fitness;
|
---|
| 217 | childScope.Fitness = cF;
|
---|
| 218 | child[idx] = !child[idx]; // undo move
|
---|
[14544] | 219 | if (Context.IsBetter(s, cF)) {
|
---|
[14420] | 220 | bestS = s;
|
---|
[14544] | 221 | bestI = i;
|
---|
[14420] | 222 | break; // first-improvement
|
---|
| 223 | }
|
---|
[14544] | 224 | if (Context.IsBetter(s, bestS)) {
|
---|
[14420] | 225 | // least-degrading
|
---|
| 226 | bestS = s;
|
---|
[14544] | 227 | bestI = i;
|
---|
[14420] | 228 | }
|
---|
| 229 | }
|
---|
[14544] | 230 | child[order[bestI]] = !child[order[bestI]];
|
---|
| 231 | order.RemoveAt(bestI);
|
---|
[14420] | 232 | cF = bestS;
|
---|
| 233 | childScope.Fitness = cF;
|
---|
[14544] | 234 | if (Context.IsBetter(cF, bF)) {
|
---|
[14420] | 235 | bF = cF;
|
---|
| 236 | best = (ISingleObjectiveSolutionScope<BinaryVector>)childScope.Clone();
|
---|
| 237 | }
|
---|
[14544] | 238 | if (order.Count == 0) break;
|
---|
[14420] | 239 | }
|
---|
[14453] | 240 | Context.IncrementEvaluatedSolutions(evaluations);
|
---|
[14420] | 241 | return best ?? childScope;
|
---|
| 242 | }
|
---|
| 243 | }
|
---|
| 244 | }
|
---|