Changeset 14559 for branches/HeuristicLab.VRPEnhancements/HeuristicLab.Problems.VehicleRouting/3.4/Encodings
- Timestamp:
- 01/12/17 16:42:50 (8 years ago)
- Location:
- branches/HeuristicLab.VRPEnhancements/HeuristicLab.Problems.VehicleRouting/3.4/Encodings/Potvin/Creators
- Files:
-
- 3 edited
Legend:
- Unmodified
- Added
- Removed
-
branches/HeuristicLab.VRPEnhancements/HeuristicLab.Problems.VehicleRouting/3.4/Encodings/Potvin/Creators/Cluster.cs
r14447 r14559 1 using System; 2 using System.Collections; 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; 3 23 using System.Collections.Generic; 4 24 using HeuristicLab.Core; 5 using HeuristicLab. Problems.VehicleRouting.Interfaces;25 using HeuristicLab.Encodings.PermutationEncoding; 6 26 7 27 namespace HeuristicLab.Problems.VehicleRouting.Encodings.Potvin { 8 28 9 public class ClusterAlgorithm<TCluster,TClusterElement> 10 where TCluster : Cluster<TClusterElement>, new() 11 where TClusterElement : ClusterElement, new() { 29 public class KMeansAlgorithm<TData> where TData : class { 12 30 13 public static List<TCluster> KMeans(IRandom random, List<TClusterElement> clusterElements, 14 int k, double changeThreshold) { 15 HashSet<int> initMeans = new HashSet<int>(); 16 int nextMean = -1; 17 List<TCluster> clusters = CreateCList(); 31 private class ClusterInfo { 32 public TData Mean; 33 public double Variance; 34 public int Size; 35 } 36 37 private readonly Func<List<TData>, TData> meanCalculator; 38 private readonly Func<TData, TData, double> distanceCalculator; 39 40 public KMeansAlgorithm(Func<List<TData>, TData> meanCalculator, Func<TData, TData, double> distanceCalculator) { 41 this.meanCalculator = meanCalculator; 42 this.distanceCalculator = distanceCalculator; 43 } 44 45 public List<List<int>> Run(List<TData> data, int k, double changeThreshold, IRandom random) { 46 int numClusters = Math.Min(k, data.Count); 47 48 var assignments = new int[data.Count]; 49 for (int i = 0; i < assignments.Length; i++) 50 assignments[i] = -1; // unassigned 18 51 19 52 // (1) initialize each cluster with a random element as mean 20 for (int i = 0; i < k && i < clusterElements.Count; i++) { 21 TCluster cluster = new TCluster(); 22 cluster.Id = i; 23 24 do { 25 nextMean = random.Next(0, clusterElements.Count); 26 } while (initMeans.Contains(nextMean)); 27 initMeans.Add(nextMean); 28 cluster.SetMean(clusterElements[nextMean]); 29 clusters.Add(cluster); 53 var clusters = new ClusterInfo[numClusters]; 54 var initIndices = new Permutation(PermutationTypes.Absolute, data.Count, random); 55 for (int c = 0; c < numClusters; c++) { 56 assignments[initIndices[c]] = c; 57 clusters[c] = new ClusterInfo { 58 Mean = data[initIndices[c]], 59 Size = 1 60 }; 30 61 } 31 62 32 // (2) repeat clustering until change rate is below th reshold33 double changeRate = 0.0;63 // (2) repeat clustering until change rate is below the threshold 64 double changeRate; 34 65 do { 35 int changes = KMeansRun(clusters, clusterElements);36 changeRate = (double)changes / clusterElements.Count;66 int changes = Iterate(data, assignments, clusters); 67 changeRate = (double)changes / data.Count; 37 68 } while (changeRate > changeThreshold); 38 69 39 40 // remove empty clusters 41 clusters.RemoveAll(c => c.Elements.Count.Equals(0)); 42 return clusters; 70 // (3) return non-empty clusters 71 var clustersData = new List<List<int>>(numClusters); 72 for (int c = 0; c < numClusters; c++) 73 clustersData.Add(new List<int>()); 74 for (int i = 0; i < assignments.Length; i++) 75 clustersData[assignments[i]].Add(i); 76 clustersData.RemoveAll(c => c.Count == 0); 77 return clustersData; 43 78 } 44 79 45 private static int KMeansRun(List<TCluster> clusters, List<TClusterElement> clusterElements) {80 private int Iterate(List<TData> data, int[] assignments, ClusterInfo[] clusters) { 46 81 int changes = 0; 47 82 48 // clear clusters from previous assigned elements 49 foreach (var c in clusters) { 50 c.Elements.Clear(); 83 var newAssignments = new int[data.Count]; 84 assignments.CopyTo(newAssignments, 0); 85 86 // assign elements to currently most suited cluster 87 for (int i = 0; i < data.Count; i++) { 88 int bestCluster = 0; 89 double bestImpact = CalculateImpact(data[i], clusters[0]); 90 for (int c = 1; c < clusters.Length; c++) { 91 double impact = CalculateImpact(data[i], clusters[c]); 92 if (impact < bestImpact) { 93 bestImpact = impact; 94 bestCluster = c; 95 } 96 } 97 newAssignments[i] = bestCluster; 98 if (newAssignments[i] != assignments[i]) 99 changes++; 51 100 } 52 101 53 // assign elements to currently most suitable clusters 54 foreach (var e in clusterElements) { 55 int optClusterIdx = 0; 56 double optImpact = clusters[optClusterIdx].CalculateImpact(e); 57 for (int i = 1; i < clusters.Count; i++) { 58 double impact = clusters[i].CalculateImpact(e); 59 if (impact < optImpact) { 60 optImpact = impact; 61 optClusterIdx = i; 62 } 63 } 64 if (clusters[optClusterIdx].AddElement(e)) { 65 changes++; 66 } 102 // update clusters 103 var clustersData = new List<List<TData>>(clusters.Length); 104 for (int c = 0; c < clusters.Length; c++) 105 clustersData.Add(new List<TData>()); 106 for (int i = 0; i < data.Count; i++) { 107 assignments[i] = newAssignments[i]; 108 clustersData[assignments[i]].Add(data[i]); 67 109 } 110 for (int c = 0; c < clusters.Length; c++) { 111 var clusterData = clustersData[c]; 112 if (clusterData.Count == 0) 113 continue; 68 114 69 // update mean and variance 70 foreach (var c in clusters) { 71 c.CalculateMean(); 72 c.CalculateVariance(); 115 clusters[c].Mean = meanCalculator(clusterData); 116 117 clusters[c].Variance = 0; 118 foreach (var e in clusterData) 119 clusters[c].Variance += Math.Pow(distanceCalculator(e, clusters[c].Mean), 2); 120 clusters[c].Variance /= clusterData.Count; 121 122 clusters[c].Size = clusterData.Count; 73 123 } 74 124 … … 76 126 } 77 127 78 private static List<TCluster> CreateCList() { 79 var listType = typeof(List<>); 80 var constructedListType = listType.MakeGenericType(typeof(TCluster)); 81 return (List<TCluster>)Activator.CreateInstance(constructedListType); 128 private double CalculateImpact(TData datum, ClusterInfo cluster) { 129 double newVariance = (cluster.Variance * cluster.Size + Math.Pow(distanceCalculator(datum, cluster.Mean), 2)) / (cluster.Size + 1); 130 return newVariance - cluster.Variance; 82 131 } 83 132 } 84 85 #region Cluster86 public interface ICluster<T> where T : ClusterElement {87 void SetMean(T o);88 bool AddElement(T o);89 void CalculateMean();90 void CalculateVariance();91 double CalculateImpact(T e);92 double CalculateDistance(T e1, T e2);93 }94 public abstract class Cluster<T> : ICluster<T> where T : ClusterElement, new() {95 public int Id;96 public List<T> Elements { get; private set; }97 public T Mean { get; set; }98 public double Variance { get; set; }99 protected Cluster() {100 Elements = new List<T>();101 }102 protected Cluster(int id) {103 Id = id;104 Elements = new List<T>();105 }106 public bool AddElement(T e) {107 Elements.Add(e);108 109 bool clusterChanged = e.ClusterId != Id;110 e.ClusterId = Id;111 return clusterChanged;112 }113 public void SetMean(T e) {114 Mean = e;115 Mean.ClusterId = Id;116 }117 public abstract void CalculateMean();118 public abstract double CalculateDistance(T e1, T e2);119 public virtual void CalculateVariance() {120 if (Mean == null)121 CalculateMean();122 123 Variance = 0.0;124 foreach (T e in Elements) {125 Variance += Math.Pow(CalculateDistance(Mean, e), 2);126 }127 Variance /= Elements.Count;128 }129 public virtual double CalculateImpact(T e) {130 if (Mean == null)131 CalculateMean();132 133 double newVariance = (Variance * Elements.Count + Math.Pow(CalculateDistance(Mean, e), 2)) / (Elements.Count + 1);134 return newVariance - Variance;135 }136 }137 public class SpatialDistanceCluster : Cluster<SpatialDistanceClusterElement> {138 public SpatialDistanceCluster() : base() {139 }140 141 public SpatialDistanceCluster(int id) : base(id) {142 }143 public override void CalculateMean() {144 int dimensions = (Mean != null) ? Mean.Coordinates.Length : (Elements.Count > 0) ? Elements[0].Coordinates.Length : 0;145 SpatialDistanceClusterElement mean = new SpatialDistanceClusterElement(-1, dimensions, Id);146 foreach (SpatialDistanceClusterElement e in Elements) {147 for (int i = 0; i < Mean.Coordinates.Length; i++) {148 mean.Coordinates[i] += (e.Coordinates[i] / Elements.Count);149 }150 }151 Mean = mean;152 }153 public override double CalculateDistance(SpatialDistanceClusterElement e1, SpatialDistanceClusterElement e2) {154 if (!e1.Coordinates.Length.Equals(e2.Coordinates.Length)) {155 throw new ArgumentException("Distance could not be calculated since number of dimensions is unequal.");156 }157 158 double distance = 0.0;159 for (int i = 0; i < e1.Coordinates.Length; i++) {160 distance += Math.Pow(e1.Coordinates[i] - e2.Coordinates[i], 2);161 }162 return Math.Sqrt(distance);163 }164 }165 public class TemporalDistanceCluster : Cluster<TemporalDistanceClusterElement> {166 public TemporalDistanceCluster() : base() {167 }168 public TemporalDistanceCluster(int id) : base(id) {169 }170 public override void CalculateMean() {171 TemporalDistanceClusterElement mean = new TemporalDistanceClusterElement(-1, Id);172 foreach (TemporalDistanceClusterElement e in Elements) {173 mean.ReadyTime += e.ReadyTime/Elements.Count;174 mean.DueTime += e.DueTime/Elements.Count;175 }176 Mean = mean;177 }178 public override double CalculateDistance(TemporalDistanceClusterElement e1, TemporalDistanceClusterElement e2) {179 double distance = 0.0;180 181 distance += Math.Pow(e1.ReadyTime - e2.ReadyTime, 2);182 distance += Math.Pow(e1.DueTime - e2.DueTime, 2);183 184 return Math.Sqrt(distance);185 }186 }187 #endregion Cluster188 189 #region ClusterElement190 public abstract class ClusterElement {191 public int Id { get; set; }192 public int ClusterId { get; set; }193 protected ClusterElement() {194 ClusterId = -1;195 }196 public ClusterElement(int id, int clusterId = -1) {197 Id = id;198 ClusterId = clusterId;199 }200 }201 public class SpatialDistanceClusterElement : ClusterElement {202 public double[] Coordinates { get; set; }203 public SpatialDistanceClusterElement() : base() {204 }205 public SpatialDistanceClusterElement(int id, int clusterId = -1) : base(id, clusterId) {206 }207 public SpatialDistanceClusterElement(int id, double[] coordinates, int clusterId = -1) : base(id, clusterId) {208 Coordinates = coordinates;209 }210 public SpatialDistanceClusterElement(int id, int dimensions, int clusterId = -1) : base(id, clusterId) {211 Coordinates = new double[dimensions];212 }213 }214 public class TemporalDistanceClusterElement : ClusterElement {215 public double ReadyTime { get; set; }216 public double DueTime { get; set; }217 public TemporalDistanceClusterElement() : base() {218 }219 public TemporalDistanceClusterElement(int id, int clusterId = -1) : base(id, clusterId) {220 }221 public TemporalDistanceClusterElement(int id, double readyTime, double dueTime, int clusterId = -1) : base(id, clusterId) {222 ReadyTime = readyTime;223 DueTime = dueTime;224 }225 }226 227 #endregion ClusterElement228 133 } -
branches/HeuristicLab.VRPEnhancements/HeuristicLab.Problems.VehicleRouting/3.4/Encodings/Potvin/Creators/GeographicDistanceClusterCreator.cs
r14442 r14559 20 20 #endregion 21 21 22 using System; 22 23 using System.Collections.Generic; 23 24 using System.Linq; … … 48 49 } 49 50 50 public static List<SpatialDistanceClusterElement> CreateClusterElements(IVRPProblemInstance instance) {51 IPickupAndDeliveryProblemInstance pdp = instance as IPickupAndDeliveryProblemInstance;52 53 // add all customers54 List<int> customers = new List<int>();55 for (int i = 1; i <= instance.Cities.Value; i++) {56 if (pdp == null || pdp.GetDemand(i) >= 0)57 customers.Add(i);58 }59 60 // wrap stops in SpatialDistanceClusterElement61 List<SpatialDistanceClusterElement> clusterElements = new List<SpatialDistanceClusterElement>();62 foreach (int customer in customers) {63 clusterElements.Add(new SpatialDistanceClusterElement(customer, instance.GetCoordinates(customer)));64 }65 return clusterElements;66 }67 51 68 52 public static PotvinEncoding CreateSolution(IVRPProblemInstance instance, IRandom random, int minK, int maxK, double clusterChangeThreshold, int creationOption) { 69 53 PotvinEncoding result = new PotvinEncoding(instance); 70 54 71 // (1) wrap stops in cluster elements 72 List<SpatialDistanceClusterElement> clusterElements = CreateClusterElements(instance); 55 // (1) init data 56 var coordinates = new List<double[]>(instance.Cities.Value); 57 var pdp = instance as IPickupAndDeliveryProblemInstance; 58 for (int i = 1; i <= instance.Cities.Value; i++) { 59 if (pdp == null || pdp.GetDemand(i) >= 0) 60 coordinates.Add(instance.GetCoordinates(i)); 61 } 73 62 74 63 // (2) create a random number k of clusters 75 64 int k = random.Next(minK, maxK); 76 List<SpatialDistanceCluster> clusters = ClusterAlgorithm<SpatialDistanceCluster, SpatialDistanceClusterElement>77 .KMeans(random, clusterElements, k, clusterChangeThreshold);78 65 var kMeans = new KMeansAlgorithm<double[]>(CalculateMeanHelper, CalculateDistanceHelper); 66 var clusters = kMeans.Run(coordinates, k, clusterChangeThreshold, random); 67 79 68 // (3) build tours with a (a) shuffling (b) greedy tour creation routine 80 foreach (var c in clusters) {69 foreach (var cluster in clusters) { 81 70 Tour newTour = new Tour(); 82 71 result.Tours.Add(newTour); … … 84 73 if (creationOption == 0) { 85 74 // (a) shuffle 86 c.Elements.Shuffle(random); 87 foreach (var o in c.Elements) { 88 newTour.Stops.Add(o.Id); 75 cluster.ShuffleInPlace(random); 76 newTour.Stops.AddRange(cluster); 77 foreach (var customer in cluster) { 78 newTour.Stops.Add(customer + 1); 89 79 } 90 80 } else { 91 81 // (b) greedy 92 foreach (var o in c.Elements) {93 newTour.Stops.Add( o.Id);82 foreach (var customer in cluster) { 83 newTour.Stops.Add(customer + 1); 94 84 } 95 85 GreedyTourCreation(instance, result, newTour, false); … … 98 88 99 89 return result; 90 } 91 92 private static double[] CalculateMeanHelper(List<double[]> coordinates) { 93 var mean = new double[coordinates[0].Length]; 94 foreach (double[] coord in coordinates) { 95 for (int i = 0; i < mean.Length; i++) { 96 mean[i] += coord[i] / coordinates.Count; 97 } 98 } 99 return mean; 100 } 101 private static double CalculateDistanceHelper(double[] coord1, double[] coord2) { 102 double distance = 0.0; 103 for (int i = 0; i < coord1.Length; i++) { 104 distance += Math.Pow(coord1[i] - coord2[i], 2); 105 } 106 return Math.Sqrt(distance); 100 107 } 101 108 -
branches/HeuristicLab.VRPEnhancements/HeuristicLab.Problems.VehicleRouting/3.4/Encodings/Potvin/Creators/TemporalDistanceClusterCreator.cs
r14442 r14559 26 26 using HeuristicLab.Core; 27 27 using HeuristicLab.Persistence.Default.CompositeSerializers.Storable; 28 using HeuristicLab.Problems.VehicleRouting.Interfaces;29 28 using HeuristicLab.Problems.VehicleRouting.Variants; 30 29 using HeuristicLab.Random; … … 49 48 } 50 49 51 public static List<TemporalDistanceClusterElement> CreateClusterElements(ITimeWindowedProblemInstance instance) {52 IPickupAndDeliveryProblemInstance pdp = instance as IPickupAndDeliveryProblemInstance;53 54 // add all customers55 List<int> customers = new List<int>();56 for (int i = 1; i <= instance.Cities.Value; i++) {57 if (pdp == null || pdp.GetDemand(i) >= 0)58 customers.Add(i);59 }60 61 // wrap stops in TemporalDistanceClusterElement62 List<TemporalDistanceClusterElement> clusterObjects = new List<TemporalDistanceClusterElement>();63 foreach (int customer in customers) {64 clusterObjects.Add(new TemporalDistanceClusterElement(customer, instance.ReadyTime[customer], instance.DueTime[customer]));65 }66 return clusterObjects;67 }68 69 50 public static PotvinEncoding CreateSolution(ITimeWindowedProblemInstance instance, IRandom random, int minK, int maxK, double clusterChangeThreshold, int creationOption) { 70 51 PotvinEncoding result = new PotvinEncoding(instance); 71 52 72 // (1) wrap stops in cluster elements 73 List<TemporalDistanceClusterElement> clusterElements = CreateClusterElements(instance); 53 // (1) init data 54 var times = new List<Tuple<double, double>>(instance.Cities.Value); 55 var pdp = instance as IPickupAndDeliveryProblemInstance; 56 for (int i = 1; i <= instance.Cities.Value; i++) { 57 if (pdp == null || pdp.GetDemand(i) >= 0) 58 times.Add(Tuple.Create(instance.ReadyTime[i], instance.DueTime[i])); 59 } 74 60 75 61 // (2) create a random number k of clusters 76 62 int k = random.Next(minK, maxK); 77 List<TemporalDistanceCluster> clusters = ClusterAlgorithm<TemporalDistanceCluster, TemporalDistanceClusterElement> 78 .KMeans(random, clusterElements, k, clusterChangeThreshold); 63 var kMeans = new KMeansAlgorithm<Tuple<double, double>>(CalculateMeanHelper, CalculateDistanceHelper); 64 var clusters = kMeans.Run(times, k, clusterChangeThreshold, random); 65 79 66 80 67 // (3) build tours with a (a) shuffling (b) greedy tour creation routine 81 foreach (var c in clusters) {68 foreach (var cluster in clusters) { 82 69 Tour newTour = new Tour(); 83 70 result.Tours.Add(newTour); … … 85 72 if (creationOption == 0) { 86 73 // (a) shuffle 87 c.Elements.Shuffle(random); 88 foreach (var o in c.Elements) { 89 newTour.Stops.Add(o.Id); 74 cluster.ShuffleInPlace(random); 75 newTour.Stops.AddRange(cluster); 76 foreach (var customer in cluster) { 77 newTour.Stops.Add(customer + 1); 90 78 } 91 79 } else { 92 80 // (b) greedy 93 foreach (var o in c.Elements) {94 newTour.Stops.Add( o.Id);81 foreach (var customer in cluster) { 82 newTour.Stops.Add(customer + 1); 95 83 } 96 84 GreedyTourCreation(instance, result, newTour, true); // "true" means include costs … … 99 87 100 88 return result; 89 } 90 91 private static Tuple<double, double> CalculateMeanHelper(List<Tuple<double, double>> times) { 92 double meanReady = 0, meanDue = 0; 93 foreach (var t in times) { 94 meanReady += t.Item1 / times.Count; 95 meanDue += t.Item2 / times.Count; 96 } 97 return Tuple.Create(meanReady, meanDue); 98 } 99 private static double CalculateDistanceHelper(Tuple<double, double> time1, Tuple<double, double> time2) { 100 double distance = 0.0; 101 distance += Math.Pow(time1.Item1 - time2.Item1, 2); 102 distance += Math.Pow(time1.Item2 - time2.Item2, 2); 103 return Math.Sqrt(distance); 101 104 } 102 105
Note: See TracChangeset
for help on using the changeset viewer.