source: branches/GeneralizedQAP/HeuristicLab.Problems.GeneralizedQuadraticAssignment/3.3/Operators/Crossovers/GQAPPathRelinking.cs @ 15558

Last change on this file since 15558 was 15558, checked in by abeham, 4 years ago

#1614: fixed bugs

File size: 13.3 KB
Line 
1#region License Information
2/* HeuristicLab
3 * Copyright (C) 2002-2017 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
22using System;
23using System.Collections.Generic;
24using System.Linq;
25using HeuristicLab.Common;
26using HeuristicLab.Core;
27using HeuristicLab.Data;
28using HeuristicLab.Encodings.IntegerVectorEncoding;
29using HeuristicLab.Parameters;
30using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
31using HeuristicLab.Random;
32
33namespace HeuristicLab.Problems.GeneralizedQuadraticAssignment {
34  /// <summary>
35  /// This is an implementation of the algorithm described in Mateus, G.R., Resende, M.G.C. & Silva, R.M.A. J Heuristics (2011) 17: 527. https://doi.org/10.1007/s10732-010-9144-0
36  /// </summary>
37  [Item("GQAPPathRelinking", "Operator that performs path relinking between two solutions. It is described in Mateus, G., Resende, M., and Silva, R. 2011. GRASP with path-relinking for the generalized quadratic assignment problem. Journal of Heuristics 17, Springer Netherlands, pp. 527-565.")]
38  [StorableClass]
39  public class GQAPPathRelinking : GQAPCrossover, IQualitiesAwareGQAPOperator {
40
41    public IScopeTreeLookupParameter<DoubleValue> QualityParameter {
42      get { return (IScopeTreeLookupParameter<DoubleValue>)Parameters["Quality"]; }
43    }
44    public IScopeTreeLookupParameter<Evaluation> EvaluationParameter {
45      get { return (IScopeTreeLookupParameter<Evaluation>)Parameters["Evaluation"]; }
46    }
47    public IValueParameter<PercentValue> CandidateSizeFactorParameter {
48      get { return (IValueParameter<PercentValue>)Parameters["CandidateSizeFactor"]; }
49    }
50    public IValueLookupParameter<BoolValue> GreedyParameter {
51      get { return (IValueLookupParameter<BoolValue>)Parameters["Greedy"]; }
52    }
53
54    [StorableConstructor]
55    protected GQAPPathRelinking(bool deserializing) : base(deserializing) { }
56    protected GQAPPathRelinking(GQAPPathRelinking original, Cloner cloner) : base(original, cloner) { }
57    public GQAPPathRelinking()
58      : base() {
59      Parameters.Add(new ScopeTreeLookupParameter<DoubleValue>("Quality", ""));
60      Parameters.Add(new ScopeTreeLookupParameter<Evaluation>("Evaluation", GQAP.EvaluationDescription));
61      Parameters.Add(new ValueParameter<PercentValue>("CandidateSizeFactor", "(η) Determines the size of the set of feasible moves in each path-relinking step relative to the maximum size. A value of 50% means that only half of all possible moves are considered each step.", new PercentValue(0.5)));
62      Parameters.Add(new ValueLookupParameter<BoolValue>("Greedy", "Whether to use a greedy selection strategy or a probabilistic one.", new BoolValue(true)));
63    }
64
65    public override IDeepCloneable Clone(Cloner cloner) {
66      return new GQAPPathRelinking(this, cloner);
67    }
68
69    public static GQAPSolution Apply(IRandom random,
70      IntegerVector source, Evaluation sourceEval,
71      IntegerVector target, Evaluation targetEval,
72      GQAPInstance problemInstance, double candidateSizeFactor,
73      out int evaluatedSolutions, bool greedy = true) {
74      evaluatedSolutions = 0;
75      var demands = problemInstance.Demands;
76      var capacities = problemInstance.Capacities;
77      var cmp = new IntegerVectorEqualityComparer();
78     
79      var sFit = problemInstance.ToSingleObjective(sourceEval);
80      var tFit = problemInstance.ToSingleObjective(targetEval);
81      GQAPSolution pi_star = sFit < tFit ? new GQAPSolution(source, sourceEval) : new GQAPSolution(target, targetEval); // line 1 of Algorithm 4
82      double pi_star_Fit = problemInstance.ToSingleObjective(pi_star.Evaluation); // line 2 of Algorithm 4
83     
84      var pi_prime = (IntegerVector)source.Clone(); // line 3 of Algorithm 4
85      //var fix = new bool[demands.Length]; // line 3 of Algorithm 4, note that according to the description it is not necessary to track the fixed equipments
86      var nonFix = Enumerable.Range(0, demands.Length).ToList(); // line 3 of Algorithm 4
87      var phi = new List<int>(IntegerVectorEqualityComparer.GetDifferingIndices(pi_prime, target)); // line 4 of Algorithm 4
88
89      var B = new List<GQAPSolution>((int)Math.Ceiling(phi.Count * candidateSizeFactor));
90      var B_fit = new List<double>(B.Capacity);
91      while (phi.Count > 0) { // line 5 of Algorithm 4
92        B.Clear(); // line 6 of Algorithm 4
93        B_fit.Clear(); // line 6 of Algorithm 4 (B is split into two synchronized lists)
94        foreach (var v in phi) { // line 7 of Algorithm 4
95          int oldLocation = pi_prime[v];
96          pi_prime[v] = target[v]; // line 8 of Algorithm 4
97          var pi_dash = MakeFeasible(random, pi_prime, v, nonFix, demands, capacities); // line 9 of Algorithm 4
98          pi_prime[v] = oldLocation; // not mentioned in Algorithm 4, but seems reasonable
99
100          if (problemInstance.IsFeasible(pi_dash)) { // line 10 of Algorithm 4
101            var pi_dash_eval = problemInstance.Evaluate(pi_dash);
102            evaluatedSolutions++;
103            var pi_dash_fit = problemInstance.ToSingleObjective(pi_dash_eval);
104
105            if (B.Any(x => cmp.Equals(x.Assignment, pi_dash))) continue; // cond. 2 of line 12 and cond. 1 of line 16 in Algorithm 4
106
107            if (B.Count >= candidateSizeFactor * phi.Count) { // line 11 of Algorithm 4
108              var replacement = B_fit.Select((val, idx) => new { Index = idx, Fitness = val })
109                                            .Where(x => x.Fitness >= pi_dash_fit) // cond. 1 in line 12 of Algorithm 4
110                                            .Select(x => new { x.Index, x.Fitness, Similarity = HammingSimilarityCalculator.CalculateSimilarity(B[x.Index].Assignment, pi_dash) })
111                                            .ToArray();
112              if (replacement.Length > 0) {
113                var mostSimilar = replacement.MaxItems(x => x.Similarity).SampleRandom(random).Index;
114                B[mostSimilar].Assignment = pi_dash; // line 13 of Algorithm 4
115                B[mostSimilar].Evaluation = pi_dash_eval; // line 13 of Algorithm 4
116                B_fit[mostSimilar] = pi_dash_fit; // line 13 of Algorithm 4
117              }
118            } else { // line 16, condition has been checked above already
119              B.Add(new GQAPSolution(pi_dash, pi_dash_eval)); // line 17 of Algorithm 4
120              B_fit.Add(pi_dash_fit); // line 17 of Algorithm 4
121            }
122          }
123        }
124        if (B.Count > 0) { // line 21 of Algorithm 4
125          GQAPSolution pi;
126          // line 22 of Algorithm 4
127          if (greedy) {
128            pi = B.Select((val, idx) => new { Index = idx, Value = val }).MinItems(x => B_fit[x.Index]).SampleRandom(random).Value;
129          } else {
130            pi = B.SampleProportional(random, 1, B_fit.Select(x => 1.0 / x), false).First();
131          }
132          var diff = IntegerVectorEqualityComparer.GetDifferingIndices(pi.Assignment, target); // line 23 of Algorithm 4
133          var I = phi.Except(diff); // line 24 of Algorithm 4
134          var i = I.SampleRandom(random); // line 25 of Algorithm 4
135          //fix[i] = true; // line 26 of Algorithm 4
136          nonFix.Remove(i); // line 26 of Algorithm 4
137          pi_prime = pi.Assignment; // line 27 of Algorithm 4
138          var fit = problemInstance.ToSingleObjective(pi.Evaluation);
139          if (fit < pi_star_Fit) { // line 28 of Algorithm 4
140            pi_star_Fit = fit; // line 29 of Algorithm 4
141            pi_star = pi; // line 30 of Algorithm 4
142          }
143        } else return pi_star;
144        phi = new List<int>(IntegerVectorEqualityComparer.GetDifferingIndices(pi_prime, target));
145      }
146
147      return pi_star;
148    }
149
150    protected override IntegerVector Cross(IRandom random, ItemArray<IntegerVector> parents,
151      GQAPInstance problemInstance) {
152
153      var qualities = QualityParameter.ActualValue;
154      var evaluations = EvaluationParameter.ActualValue;
155      var betterParent = qualities[0].Value <= qualities[1].Value ? 0 : 1;
156      var worseParent = 1 - betterParent;
157      var source = parents[betterParent];
158      var target = parents[worseParent];
159
160      int evaluatedSolution;
161      return Apply(random, source, evaluations[betterParent],
162        target, evaluations[worseParent], problemInstance,
163        CandidateSizeFactorParameter.Value.Value, out evaluatedSolution,
164        GreedyParameter.ActualValue.Value).Assignment;
165    }
166
167    /// <summary>
168    /// Relocates equipments in the same location as <paramref name="equipment"/> to other locations in case the location
169    /// is overutilized.
170    /// </summary>
171    /// <remarks>
172    /// This method is performance critical, called very often and should run as fast as possible.
173    /// </remarks>
174    /// <param name="random">The random number generator.</param>
175    /// <param name="pi">The current solution.</param>
176    /// <param name="equipment">The equipment that was just assigned to a new location.</param>
177    /// <param name="nonFix">The equipments that have not yet been fixed.</param>
178    /// <param name="demands">The demands for all equipments.</param>
179    /// <param name="capacities">The capacities of all locations.</param>
180    /// <param name="maximumTries">The number of tries that should be done in relocating the equipments.</param>
181    /// <returns>A feasible or infeasible solution</returns>
182    private static IntegerVector MakeFeasible(IRandom random, IntegerVector pi, int equipment, List<int> nonFix, DoubleArray demands, DoubleArray capacities, int maximumTries = 1000) {
183      int l = pi[equipment];
184      var slack = ComputeSlack(pi, demands, capacities);
185      if (slack[l] >= 0) // line 1 of Algorithm 5
186        return (IntegerVector)pi.Clone(); // line 2 of Algorithm 5
187
188      IntegerVector pi_prime = null;
189      int k = 0; // line 4 of Algorithm 5
190      var maxSlack = slack.Max(); // line 8-9 of Algorithm 5
191      var slack_prime = (double[])slack.Clone();
192      var maxSlack_prime = maxSlack;
193      // note that FTL can be computed only once for all tries as all tries restart with the same solution
194      var FTL = nonFix.Where(x => x != equipment && pi[x] == l && demands[x] <= maxSlack).ToList(); // line 8-9 of Algorithm 5
195      var FTLweight = FTL.Select(x => demands[x]).ToList();
196      while (k < maximumTries && slack_prime[l] < 0) {  // line 5 of Algorithm 5
197        pi_prime = (IntegerVector)pi.Clone(); // line 6 of Algorithm 5
198        // set T can only shrink and not grow, thus it is created outside the loop and only updated inside
199        var T = new List<int>(FTL); // line 8-9 of Algorithm 5
200        var weightT = new List<double>(FTLweight);
201        do {  // line 7 of Algorithm 5
202          if (T.Count > 0) { // line 10 of Algorithm 5
203            var idx = Enumerable.Range(0, T.Count).SampleProportional(random, 1, weightT, false, false).First(); // line 11 of Algorithm 5
204            int i = T[idx]; // line 11 of Algorithm 5
205            var j = Enumerable.Range(0, capacities.Length)
206              .Where(x => slack_prime[x] >= demands[i]) // line 12 of Algorithm 5
207              .SampleRandom(random);  // line 13 of Algorithm 5
208            pi_prime[i] = j; // line 14 of Algorithm 5
209            T.RemoveAt(idx);
210            weightT.RemoveAt(idx);
211            var recomputeMaxSlack = slack_prime[j] == maxSlack_prime; // efficiency improvement: recompute max slack only if we assign to a location whose slack equals maxSlack
212            slack_prime[j] -= demands[i]; // line 14 of Algorithm 5
213            slack_prime[l] += demands[i]; // line 14 of Algorithm 5
214            if (recomputeMaxSlack) {
215              maxSlack_prime = slack_prime.Max();
216              // T needs to be removed of equipments whose demand is higher than maxSlack only if maxSlack changes
217              for (var h = 0; h < T.Count; h++) {
218                var f = T[h];
219                if (demands[f] > maxSlack_prime) {
220                  T.RemoveAt(h);
221                  weightT.RemoveAt(h);
222                  h--;
223                }
224              }
225            }
226          } else break; // cond. 1 in line 16 of Algorithm 5
227        } while (slack_prime[l] < 0); // cond. 2 in line 16 of Algorithm 5
228        k++; // line 17 of Algorithm 5
229        if (slack_prime[l] < 0) {
230          // reset
231          Array.Copy(slack, slack_prime, slack.Length);
232          maxSlack_prime = maxSlack;
233        }
234      }
235      return pi_prime; // line 19-23 of Algorithm 5
236    }
237
238    private static double[] ComputeSlack(IntegerVector assignment, DoubleArray demands, DoubleArray capacities) {
239      var slack = capacities.ToArray();
240      for (int i = 0; i < assignment.Length; i++) {
241        slack[assignment[i]] -= demands[i];
242      }
243      return slack;
244    }
245  }
246}
Note: See TracBrowser for help on using the repository browser.