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 | * Implementation is based on jMetal framework https://github.com/jMetal/jMetal
|
---|
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 | #endregion
|
---|
22 | using HeuristicLab.Analysis;
|
---|
23 | using HeuristicLab.Common;
|
---|
24 | using HeuristicLab.Core;
|
---|
25 | using HeuristicLab.Data;
|
---|
26 | using HeuristicLab.Encodings.RealVectorEncoding;
|
---|
27 | using HeuristicLab.Optimization;
|
---|
28 | using HeuristicLab.Parameters;
|
---|
29 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
30 | using HeuristicLab.Problems.TestFunctions;
|
---|
31 | using HeuristicLab.Random;
|
---|
32 | using System;
|
---|
33 | using System.Collections.Generic;
|
---|
34 | using System.Threading;
|
---|
35 |
|
---|
36 | namespace HeuristicLab.Algorithms.DifferentialEvolution
|
---|
37 | {
|
---|
38 |
|
---|
39 | [Item("Differential Evolution (DE)", "A differential evolution algorithm.")]
|
---|
40 | [StorableClass]
|
---|
41 | [Creatable(CreatableAttribute.Categories.PopulationBasedAlgorithms, Priority = 400)]
|
---|
42 | public class DifferentialEvolution : BasicAlgorithm
|
---|
43 | {
|
---|
44 | public Func<IEnumerable<double>, double> Evaluation;
|
---|
45 |
|
---|
46 | public override Type ProblemType
|
---|
47 | {
|
---|
48 | get { return typeof(SingleObjectiveTestFunctionProblem); }
|
---|
49 | }
|
---|
50 | public new SingleObjectiveTestFunctionProblem Problem
|
---|
51 | {
|
---|
52 | get { return (SingleObjectiveTestFunctionProblem)base.Problem; }
|
---|
53 | set { base.Problem = value; }
|
---|
54 | }
|
---|
55 |
|
---|
56 | private readonly IRandom _random = new MersenneTwister();
|
---|
57 | private int evals;
|
---|
58 |
|
---|
59 | #region ParameterNames
|
---|
60 | private const string MaximumEvaluationsParameterName = "Maximum Evaluations";
|
---|
61 | private const string SeedParameterName = "Seed";
|
---|
62 | private const string SetSeedRandomlyParameterName = "SetSeedRandomly";
|
---|
63 | private const string CrossoverProbabilityParameterName = "CrossoverProbability";
|
---|
64 | private const string PopulationSizeParameterName = "PopulationSize";
|
---|
65 | private const string ScalingFactorParameterName = "ScalingFactor";
|
---|
66 | private const string ValueToReachParameterName = "ValueToReach";
|
---|
67 | #endregion
|
---|
68 |
|
---|
69 | #region ParameterProperties
|
---|
70 | public IFixedValueParameter<IntValue> MaximumEvaluationsParameter
|
---|
71 | {
|
---|
72 | get { return (IFixedValueParameter<IntValue>)Parameters[MaximumEvaluationsParameterName]; }
|
---|
73 | }
|
---|
74 | public IFixedValueParameter<IntValue> SeedParameter
|
---|
75 | {
|
---|
76 | get { return (IFixedValueParameter<IntValue>)Parameters[SeedParameterName]; }
|
---|
77 | }
|
---|
78 | public FixedValueParameter<BoolValue> SetSeedRandomlyParameter
|
---|
79 | {
|
---|
80 | get { return (FixedValueParameter<BoolValue>)Parameters[SetSeedRandomlyParameterName]; }
|
---|
81 | }
|
---|
82 | private ValueParameter<IntValue> PopulationSizeParameter
|
---|
83 | {
|
---|
84 | get { return (ValueParameter<IntValue>)Parameters[PopulationSizeParameterName]; }
|
---|
85 | }
|
---|
86 | public ValueParameter<DoubleValue> CrossoverProbabilityParameter
|
---|
87 | {
|
---|
88 | get { return (ValueParameter<DoubleValue>)Parameters[CrossoverProbabilityParameterName]; }
|
---|
89 | }
|
---|
90 | public ValueParameter<DoubleValue> ScalingFactorParameter
|
---|
91 | {
|
---|
92 | get { return (ValueParameter<DoubleValue>)Parameters[ScalingFactorParameterName]; }
|
---|
93 | }
|
---|
94 | public ValueParameter<DoubleValue> ValueToReachParameter
|
---|
95 | {
|
---|
96 | get { return (ValueParameter<DoubleValue>)Parameters[ValueToReachParameterName]; }
|
---|
97 | }
|
---|
98 | #endregion
|
---|
99 |
|
---|
100 | #region Properties
|
---|
101 | public int MaximumEvaluations
|
---|
102 | {
|
---|
103 | get { return MaximumEvaluationsParameter.Value.Value; }
|
---|
104 | set { MaximumEvaluationsParameter.Value.Value = value; }
|
---|
105 | }
|
---|
106 |
|
---|
107 | public Double CrossoverProbability
|
---|
108 | {
|
---|
109 | get { return CrossoverProbabilityParameter.Value.Value; }
|
---|
110 | set { CrossoverProbabilityParameter.Value.Value = value; }
|
---|
111 | }
|
---|
112 | public Double ScalingFactor
|
---|
113 | {
|
---|
114 | get { return ScalingFactorParameter.Value.Value; }
|
---|
115 | set { ScalingFactorParameter.Value.Value = value; }
|
---|
116 | }
|
---|
117 | public int Seed
|
---|
118 | {
|
---|
119 | get { return SeedParameter.Value.Value; }
|
---|
120 | set { SeedParameter.Value.Value = value; }
|
---|
121 | }
|
---|
122 | public bool SetSeedRandomly
|
---|
123 | {
|
---|
124 | get { return SetSeedRandomlyParameter.Value.Value; }
|
---|
125 | set { SetSeedRandomlyParameter.Value.Value = value; }
|
---|
126 | }
|
---|
127 | public IntValue PopulationSize
|
---|
128 | {
|
---|
129 | get { return PopulationSizeParameter.Value; }
|
---|
130 | set { PopulationSizeParameter.Value = value; }
|
---|
131 | }
|
---|
132 | public Double ValueToReach
|
---|
133 | {
|
---|
134 | get { return ValueToReachParameter.Value.Value; }
|
---|
135 | set { ValueToReachParameter.Value.Value = value; }
|
---|
136 | }
|
---|
137 | #endregion
|
---|
138 |
|
---|
139 | #region ResultsProperties
|
---|
140 | private double ResultsBestQuality
|
---|
141 | {
|
---|
142 | get { return ((DoubleValue)Results["Best Quality"].Value).Value; }
|
---|
143 | set { ((DoubleValue)Results["Best Quality"].Value).Value = value; }
|
---|
144 | }
|
---|
145 |
|
---|
146 | private double VTRBestQuality
|
---|
147 | {
|
---|
148 | get { return ((DoubleValue)Results["VTR"].Value).Value; }
|
---|
149 | set { ((DoubleValue)Results["VTR"].Value).Value = value; }
|
---|
150 | }
|
---|
151 |
|
---|
152 | private RealVector ResultsBestSolution
|
---|
153 | {
|
---|
154 | get { return (RealVector)Results["Best Solution"].Value; }
|
---|
155 | set { Results["Best Solution"].Value = value; }
|
---|
156 | }
|
---|
157 |
|
---|
158 | private int ResultsEvaluations
|
---|
159 | {
|
---|
160 | get { return ((IntValue)Results["Evaluations"].Value).Value; }
|
---|
161 | set { ((IntValue)Results["Evaluations"].Value).Value = value; }
|
---|
162 | }
|
---|
163 | private int ResultsIterations
|
---|
164 | {
|
---|
165 | get { return ((IntValue)Results["Iterations"].Value).Value; }
|
---|
166 | set { ((IntValue)Results["Iterations"].Value).Value = value; }
|
---|
167 | }
|
---|
168 |
|
---|
169 | private DataTable ResultsQualities
|
---|
170 | {
|
---|
171 | get { return ((DataTable)Results["Qualities"].Value); }
|
---|
172 | }
|
---|
173 | private DataRow ResultsQualitiesBest
|
---|
174 | {
|
---|
175 | get { return ResultsQualities.Rows["Best Quality"]; }
|
---|
176 | }
|
---|
177 |
|
---|
178 | #endregion
|
---|
179 |
|
---|
180 | [StorableConstructor]
|
---|
181 | protected DifferentialEvolution(bool deserializing) : base(deserializing) { }
|
---|
182 |
|
---|
183 | protected DifferentialEvolution(DifferentialEvolution original, Cloner cloner)
|
---|
184 | : base(original, cloner)
|
---|
185 | {
|
---|
186 | }
|
---|
187 |
|
---|
188 | public override IDeepCloneable Clone(Cloner cloner)
|
---|
189 | {
|
---|
190 | return new DifferentialEvolution(this, cloner);
|
---|
191 | }
|
---|
192 |
|
---|
193 | public DifferentialEvolution()
|
---|
194 | {
|
---|
195 | Parameters.Add(new FixedValueParameter<IntValue>(MaximumEvaluationsParameterName, "", new IntValue(Int32.MaxValue)));
|
---|
196 | Parameters.Add(new FixedValueParameter<IntValue>(SeedParameterName, "The random seed used to initialize the new pseudo random number generator.", new IntValue(0)));
|
---|
197 | Parameters.Add(new FixedValueParameter<BoolValue>(SetSeedRandomlyParameterName, "True if the random seed should be set to a random value, otherwise false.", new BoolValue(true)));
|
---|
198 | Parameters.Add(new ValueParameter<IntValue>(PopulationSizeParameterName, "The size of the population of solutions.", new IntValue(100)));
|
---|
199 | Parameters.Add(new ValueParameter<DoubleValue>(CrossoverProbabilityParameterName, "The value for crossover rate", new DoubleValue(0.88)));
|
---|
200 | Parameters.Add(new ValueParameter<DoubleValue>(ScalingFactorParameterName, "The value for scaling factor", new DoubleValue(0.47)));
|
---|
201 | Parameters.Add(new ValueParameter<DoubleValue>(ValueToReachParameterName, "Value to reach (VTR) parameter", new DoubleValue(0.00000001)));
|
---|
202 | }
|
---|
203 |
|
---|
204 | protected override void Run(CancellationToken cancellationToken)
|
---|
205 | {
|
---|
206 |
|
---|
207 | // Set up the results display
|
---|
208 | Results.Add(new Result("Iterations", new IntValue(0)));
|
---|
209 | Results.Add(new Result("Evaluations", new IntValue(0)));
|
---|
210 | Results.Add(new Result("Best Solution", new RealVector()));
|
---|
211 | Results.Add(new Result("Best Quality", new DoubleValue(double.NaN)));
|
---|
212 | Results.Add(new Result("VTR", new DoubleValue(double.NaN)));
|
---|
213 | var table = new DataTable("Qualities");
|
---|
214 | table.Rows.Add(new DataRow("Best Quality"));
|
---|
215 | Results.Add(new Result("Qualities", table));
|
---|
216 |
|
---|
217 |
|
---|
218 | //problem variables
|
---|
219 | var dim = Problem.ProblemSize.Value;
|
---|
220 | var lb = Problem.Bounds[0, 0];
|
---|
221 | var ub = Problem.Bounds[0, 1];
|
---|
222 | var range = ub - lb;
|
---|
223 | this.evals = 0;
|
---|
224 |
|
---|
225 | double[,] populationOld = new double[PopulationSizeParameter.Value.Value, Problem.ProblemSize.Value];
|
---|
226 | double[,] mutationPopulation = new double[PopulationSizeParameter.Value.Value, Problem.ProblemSize.Value];
|
---|
227 | double[,] trialPopulation = new double[PopulationSizeParameter.Value.Value, Problem.ProblemSize.Value];
|
---|
228 | double[] bestPopulation = new double[Problem.ProblemSize.Value];
|
---|
229 | double[] bestPopulationIteration = new double[Problem.ProblemSize.Value];
|
---|
230 |
|
---|
231 | //create initial population
|
---|
232 | //population is a matrix of size PopulationSize*ProblemSize
|
---|
233 | for (int i = 0; i < PopulationSizeParameter.Value.Value; i++)
|
---|
234 | {
|
---|
235 | for (int j = 0; j < Problem.ProblemSize.Value; j++)
|
---|
236 | {
|
---|
237 | populationOld[i, j] = _random.NextDouble() * range + lb;
|
---|
238 | }
|
---|
239 | }
|
---|
240 |
|
---|
241 | //evaluate the best member after the intialiazation
|
---|
242 | //the idea is to select first member and after that to check the others members from the population
|
---|
243 |
|
---|
244 | int best_index = 0;
|
---|
245 | double[] populationRow = new double[Problem.ProblemSize.Value];
|
---|
246 | double[] qualityPopulation = new double[PopulationSizeParameter.Value.Value];
|
---|
247 | bestPopulation = getMatrixRow(populationOld, best_index);
|
---|
248 | RealVector bestPopulationVector = new RealVector(bestPopulation);
|
---|
249 | double bestPopulationValue = Obj(bestPopulationVector);
|
---|
250 | qualityPopulation[best_index] = bestPopulationValue;
|
---|
251 | RealVector selectionVector;
|
---|
252 | RealVector trialVector;
|
---|
253 | double qtrial;
|
---|
254 |
|
---|
255 |
|
---|
256 | for (var i = 1; i < PopulationSizeParameter.Value.Value; i++)
|
---|
257 | {
|
---|
258 | populationRow = getMatrixRow(populationOld, i);
|
---|
259 | trialVector = new RealVector(populationRow);
|
---|
260 |
|
---|
261 | qtrial = Obj(trialVector);
|
---|
262 | qualityPopulation[i] = qtrial;
|
---|
263 |
|
---|
264 | if (qtrial > bestPopulationValue)
|
---|
265 | {
|
---|
266 | bestPopulationVector = new RealVector(populationRow);
|
---|
267 | bestPopulationValue = qtrial;
|
---|
268 | best_index = i;
|
---|
269 | }
|
---|
270 | }
|
---|
271 |
|
---|
272 | int iterations = 1;
|
---|
273 |
|
---|
274 | // Loop until iteration limit reached or canceled.
|
---|
275 | // todo replace with a function
|
---|
276 | // && bestPopulationValue > Problem.BestKnownQuality.Value + ValueToReachParameter.Value.Value
|
---|
277 | while (ResultsEvaluations < MaximumEvaluations
|
---|
278 | && !cancellationToken.IsCancellationRequested
|
---|
279 | && (bestPopulationValue - Problem.BestKnownQuality.Value) > ValueToReachParameter.Value.Value)
|
---|
280 | {
|
---|
281 | //mutation DE/rand/1/bin; classic DE
|
---|
282 | for (int i = 0; i < PopulationSizeParameter.Value.Value; i++)
|
---|
283 | {
|
---|
284 | int r0, r1, r2;
|
---|
285 |
|
---|
286 | //assure the selected vectors r0, r1 and r2 are different
|
---|
287 | do
|
---|
288 | {
|
---|
289 | r0 = _random.Next(0, PopulationSizeParameter.Value.Value);
|
---|
290 | } while (r0 == i);
|
---|
291 | do
|
---|
292 | {
|
---|
293 | r1 = _random.Next(0, PopulationSizeParameter.Value.Value);
|
---|
294 | } while (r1 == i || r1 == r0);
|
---|
295 | do
|
---|
296 | {
|
---|
297 | r2 = _random.Next(0, PopulationSizeParameter.Value.Value);
|
---|
298 | } while (r2 == i || r2 == r0 || r2 == r1);
|
---|
299 |
|
---|
300 | for (int j = 0; j < getMatrixRow(mutationPopulation, i).Length; j++)
|
---|
301 | {
|
---|
302 | mutationPopulation[i, j] = populationOld[r0, j] +
|
---|
303 | ScalingFactorParameter.Value.Value * (populationOld[r1, j] - populationOld[r2, j]);
|
---|
304 | //check the problem upper and lower bounds
|
---|
305 | if (mutationPopulation[i, j] > ub) mutationPopulation[i, j] = ub;
|
---|
306 | if (mutationPopulation[i, j] < lb) mutationPopulation[i, j] = lb;
|
---|
307 | }
|
---|
308 | }
|
---|
309 |
|
---|
310 | //uniform crossover
|
---|
311 | for (int i = 0; i < PopulationSizeParameter.Value.Value; i++)
|
---|
312 | {
|
---|
313 | double rnbr = _random.Next(0, Problem.ProblemSize.Value);
|
---|
314 | for (int j = 0; j < getMatrixRow(mutationPopulation, i).Length; j++)
|
---|
315 | {
|
---|
316 | if (_random.NextDouble() <= CrossoverProbabilityParameter.Value.Value || j == rnbr)
|
---|
317 | {
|
---|
318 | trialPopulation[i, j] = mutationPopulation[i, j];
|
---|
319 | }
|
---|
320 | else
|
---|
321 | {
|
---|
322 | trialPopulation[i, j] = populationOld[i, j];
|
---|
323 | }
|
---|
324 | }
|
---|
325 | }
|
---|
326 |
|
---|
327 | //One-to-One Survivor Selection
|
---|
328 | for (int i = 0; i < PopulationSizeParameter.Value.Value; i++)
|
---|
329 | {
|
---|
330 | selectionVector = new RealVector(getMatrixRow(populationOld, i));
|
---|
331 | trialVector = new RealVector(getMatrixRow(trialPopulation, i));
|
---|
332 |
|
---|
333 | var selectionEval = qualityPopulation[i];
|
---|
334 | var trialEval = Obj(trialVector);
|
---|
335 |
|
---|
336 | if (trialEval < selectionEval)
|
---|
337 | {
|
---|
338 | for (int j = 0; j < getMatrixRow(populationOld, i).Length; j++)
|
---|
339 | {
|
---|
340 | populationOld[i, j] = trialPopulation[i, j];
|
---|
341 | }
|
---|
342 | qualityPopulation[i] = trialEval;
|
---|
343 | }
|
---|
344 | }
|
---|
345 |
|
---|
346 | //update the best candidate
|
---|
347 | for (int i = 0; i < PopulationSizeParameter.Value.Value; i++)
|
---|
348 | {
|
---|
349 | selectionVector = new RealVector(getMatrixRow(populationOld, i));
|
---|
350 | var quality = qualityPopulation[i];
|
---|
351 | if (quality < bestPopulationValue)
|
---|
352 | {
|
---|
353 | bestPopulationVector = (RealVector)selectionVector.Clone();
|
---|
354 | bestPopulationValue = quality;
|
---|
355 | }
|
---|
356 | }
|
---|
357 |
|
---|
358 | iterations = iterations + 1;
|
---|
359 |
|
---|
360 | //update the results
|
---|
361 | ResultsEvaluations = evals;
|
---|
362 | ResultsIterations = iterations;
|
---|
363 | ResultsBestSolution = bestPopulationVector;
|
---|
364 | ResultsBestQuality = bestPopulationValue;
|
---|
365 |
|
---|
366 | //update the results in view
|
---|
367 | if (iterations % 10 == 0) ResultsQualitiesBest.Values.Add(bestPopulationValue);
|
---|
368 | if (bestPopulationValue < Problem.BestKnownQuality.Value + ValueToReachParameter.Value.Value)
|
---|
369 | {
|
---|
370 | VTRBestQuality = bestPopulationValue;
|
---|
371 | }
|
---|
372 | }
|
---|
373 | }
|
---|
374 |
|
---|
375 | //evaluate the vector
|
---|
376 | public double Obj(RealVector x)
|
---|
377 | {
|
---|
378 | evals = evals + 1;
|
---|
379 | if (Problem.Maximization.Value)
|
---|
380 | return -Problem.Evaluator.Evaluate(x);
|
---|
381 |
|
---|
382 | return Problem.Evaluator.Evaluate(x);
|
---|
383 | }
|
---|
384 |
|
---|
385 | // Get ith row from the matrix
|
---|
386 | public double[] getMatrixRow(double[,] Mat, int i)
|
---|
387 | {
|
---|
388 | double[] tmp = new double[Mat.GetUpperBound(1) + 1];
|
---|
389 |
|
---|
390 | for (int j = 0; j <= Mat.GetUpperBound(1); j++)
|
---|
391 | {
|
---|
392 | tmp[j] = Mat[i, j];
|
---|
393 | }
|
---|
394 |
|
---|
395 | return tmp;
|
---|
396 | }
|
---|
397 | }
|
---|
398 | }
|
---|