[14678] | 1 | using System;
|
---|
| 2 | using System.Collections.Generic;
|
---|
[14690] | 3 | using System.Globalization;
|
---|
[14678] | 4 | using System.Linq;
|
---|
| 5 | using System.Threading.Tasks;
|
---|
| 6 | using HeuristicLab.PluginInfrastructure;
|
---|
| 7 | using HeuristicLab.Problems.Instances;
|
---|
| 8 | using HeuristicLab.Problems.Instances.QAPLIB;
|
---|
| 9 | using HeuristicLab.Problems.QuadraticAssignment;
|
---|
[14691] | 10 | using HeuristicLab.Random;
|
---|
[14678] | 11 |
|
---|
| 12 | namespace ProblemInstanceIdentifier {
|
---|
| 13 | class Program {
|
---|
[14690] | 14 | private static HashSet<string> selectedInstances = new HashSet<string>() {
|
---|
| 15 | "bur26a", "chr25a", "dre24", "els19", "esc32a", "had20", "kra32", "lipa30a", "lipa30b",
|
---|
| 16 | "nug30", "rou20", "scr20", "sko42", "ste36a", "tai25a", "tai25b", "tho30", "wil50",
|
---|
| 17 | "RAND-S-6x6-36-25-AFFY-00_rand_rand_bl", "RAND-S-6x6-36-25-AFFY-00_rand_rand_ci"
|
---|
| 18 | };
|
---|
[14678] | 19 | static void Main(string[] args) {
|
---|
[14691] | 20 | var instances = Get20DifferentClasses();
|
---|
| 21 | //var instances = GetSomeRandomInstances(50);
|
---|
| 22 |
|
---|
| 23 | /*var classes = instances.Select(InstanceDescriptor.FromProblemOnly).GroupBy(x => x.Cls);
|
---|
| 24 | foreach (var cls in classes.OrderBy(x => x.Key)) {
|
---|
| 25 | Console.WriteLine("{0};{1}", cls.Key, cls.Count());
|
---|
| 26 | }*/
|
---|
| 27 |
|
---|
| 28 | var prExplorer = new PathRelinkingExplorer() {
|
---|
| 29 | LocalOptima = false,
|
---|
| 30 | Paths = 200
|
---|
| 31 | };
|
---|
| 32 | var prLocalExplorer = new PathRelinkingExplorer() {
|
---|
| 33 | LocalOptima = true,
|
---|
| 34 | Paths = 200
|
---|
| 35 | };
|
---|
| 36 | var memPrExplorer = new MemPRExplorer() {
|
---|
| 37 | IncludeLocalSearch = false,
|
---|
| 38 | Seconds = 10
|
---|
| 39 | };
|
---|
| 40 |
|
---|
| 41 | var training = GenerateData(instances, prExplorer, parallel:true);
|
---|
| 42 | var standardizer = InstancesStandardizer.CreateAndApply(training);
|
---|
| 43 | var test = GenerateData(instances, prExplorer, parallel: false);
|
---|
| 44 | standardizer.Apply(test);
|
---|
| 45 | PrintMatchResult(Compare(training, test));
|
---|
| 46 |
|
---|
| 47 | ExploreMatching(instances, new [] { prLocalExplorer }, new [] { memPrExplorer });
|
---|
| 48 | }
|
---|
| 49 |
|
---|
| 50 | private static List<QuadraticAssignmentProblem> GetSomeRandomInstances(int totalInstances) {
|
---|
[14678] | 51 | var sync = new object();
|
---|
[14691] | 52 | var provider = new OneSizeInstanceProvider();
|
---|
| 53 | var instances = new List<QuadraticAssignmentProblem>();
|
---|
| 54 | var random = new FastRandom(0);
|
---|
| 55 | Parallel.ForEach(provider.GetDataDescriptors().Shuffle(random), desc => {
|
---|
| 56 | var qapData = provider.LoadData(desc);
|
---|
| 57 | if (qapData.Dimension < 25) return;
|
---|
| 58 | if (instances.Count >= totalInstances) return;
|
---|
| 59 | var qap = new QuadraticAssignmentProblem();
|
---|
| 60 | qap.Load(qapData);
|
---|
| 61 | lock (sync) {
|
---|
| 62 | if (instances.Count >= totalInstances) return;
|
---|
| 63 | instances.Add(qap);
|
---|
| 64 | }
|
---|
| 65 | });
|
---|
| 66 | return instances;
|
---|
| 67 | }
|
---|
[14678] | 68 |
|
---|
[14691] | 69 | private static List<QuadraticAssignmentProblem> Get20DifferentClasses() {
|
---|
| 70 | var sync = new object();
|
---|
| 71 |
|
---|
[14678] | 72 | var qapProviders = ApplicationManager.Manager.GetInstances<ProblemInstanceProvider<QAPData>>().ToList();
|
---|
| 73 | var instances = new List<QuadraticAssignmentProblem>();
|
---|
| 74 | foreach (var provider in qapProviders) {
|
---|
| 75 | if (provider is TaillardQAPInstanceProvider) continue;
|
---|
| 76 | Parallel.ForEach(provider.GetDataDescriptors(), desc => {
|
---|
[14690] | 77 | if (!selectedInstances.Contains(desc.Name)) return;
|
---|
| 78 | //if (desc.Name == "esc16f") return;
|
---|
[14678] | 79 | var qapData = provider.LoadData(desc);
|
---|
[14690] | 80 | //if (qapData.Dimension < 15 || qapData.Dimension > 150) return;
|
---|
[14678] | 81 | var qap = new QuadraticAssignmentProblem();
|
---|
| 82 | qap.Load(qapData);
|
---|
| 83 | lock (sync) {
|
---|
| 84 | instances.Add(qap);
|
---|
| 85 | }
|
---|
| 86 | });
|
---|
| 87 | }
|
---|
[14691] | 88 | return instances;
|
---|
[14690] | 89 | }
|
---|
| 90 |
|
---|
[14691] | 91 | private static List<InstanceDescriptor> GenerateData(List<QuadraticAssignmentProblem> instances, InstanceExplorer explorer, bool parallel = false) {
|
---|
[14690] | 92 | var sync = new object();
|
---|
[14691] | 93 | var data = new List<InstanceDescriptor>();
|
---|
| 94 | Action<QuadraticAssignmentProblem> body = (qap) => {
|
---|
[14690] | 95 | var instance = explorer.Explore(qap);
|
---|
[14691] | 96 | if (instance == null) return;
|
---|
[14690] | 97 | lock (sync) {
|
---|
[14691] | 98 | data.Add(instance);
|
---|
[14690] | 99 | }
|
---|
[14691] | 100 | };
|
---|
| 101 | if (parallel) {
|
---|
| 102 | Parallel.ForEach(instances, body);
|
---|
| 103 | } else {
|
---|
| 104 | foreach (var qap in instances) body(qap);
|
---|
[14690] | 105 | }
|
---|
[14691] | 106 | return data;
|
---|
[14690] | 107 | }
|
---|
| 108 |
|
---|
[14691] | 109 | private static MatchResult Compare(List<InstanceDescriptor> training, List<InstanceDescriptor> test) {
|
---|
| 110 | int exactCount = 0, clsCount = 0, totalCount = 0;
|
---|
| 111 | int exactRank = 0, clsRank = 0;
|
---|
| 112 | foreach (var e in test) {
|
---|
| 113 | var ordered = training.OrderBy(x => x.CalculateSimilarity(e)).ToList();
|
---|
| 114 | var bestMatch = ordered.First();
|
---|
| 115 | if (bestMatch.Cls == e.Cls) clsCount++;
|
---|
| 116 | if (bestMatch.Name == e.Name) exactCount++;
|
---|
[14696] | 117 | var r = ordered.FindIndex((id) => id.Name == e.Name);
|
---|
| 118 | if (r < 0) continue;
|
---|
[14691] | 119 | totalCount++;
|
---|
[14696] | 120 | exactRank += r + 1;
|
---|
[14691] | 121 | clsRank += ordered.FindIndex((id) => id.Cls == e.Cls) + 1;
|
---|
| 122 | }
|
---|
| 123 |
|
---|
| 124 | return new MatchResult() {
|
---|
| 125 | ExactCount = exactCount,
|
---|
| 126 | ClsCount = clsCount,
|
---|
| 127 | TotalCount = totalCount,
|
---|
| 128 | ExactAverageRank = exactRank / (double)totalCount,
|
---|
| 129 | ClsAverageRank = clsRank / (double)totalCount
|
---|
| 130 | };
|
---|
| 131 | }
|
---|
| 132 |
|
---|
| 133 | private static void PrintMatchResult(MatchResult result) {
|
---|
| 134 | Console.WriteLine("{0}\t{1}\t{2}\t{3:F2}\t{4:F2}",
|
---|
| 135 | result.ExactCount, result.ClsCount, result.TotalCount,
|
---|
| 136 | result.ExactAverageRank, result.ClsAverageRank);
|
---|
| 137 | }
|
---|
| 138 |
|
---|
| 139 | private static void PrintData(List<InstanceDescriptor> instances) {
|
---|
| 140 | using (var iter = instances.GetEnumerator()) {
|
---|
| 141 | if (!iter.MoveNext()) return;
|
---|
| 142 | Console.WriteLine(string.Join(";", new[] {"Name", "Cls", "Dimension"}
|
---|
| 143 | .Concat(iter.Current != null ? iter.Current.FeatureNames : new [] { "(null)" })));
|
---|
| 144 | do {
|
---|
| 145 | PrintInstanceLine(iter.Current);
|
---|
| 146 | } while (iter.MoveNext());
|
---|
| 147 | }
|
---|
| 148 | }
|
---|
| 149 |
|
---|
[14690] | 150 | private static void PrintInstanceLine(InstanceDescriptor instance) {
|
---|
| 151 | Console.WriteLine(string.Join(";",
|
---|
| 152 | new[] {instance.Name, instance.Cls, instance.Dimension.ToString(CultureInfo.CurrentCulture)}
|
---|
| 153 | .Concat(instance.FeatureValues.Select(x => x.ToString(CultureInfo.CurrentCulture)))));
|
---|
| 154 | }
|
---|
| 155 |
|
---|
[14691] | 156 | private static void ExploreMatching(List<QuadraticAssignmentProblem> instances, InstanceExplorer[] trainingExplorers, InstanceExplorer[] testExporers, bool parallel = false) {
|
---|
[14690] | 157 | var sync = new object();
|
---|
[14691] | 158 | var rand = new Random();
|
---|
| 159 | var first = rand.Next();
|
---|
| 160 | var second = rand.Next();
|
---|
| 161 | while (first == second) second = rand.Next();
|
---|
| 162 |
|
---|
| 163 | var knowledgeBase = new Dictionary<InstanceExplorer, Tuple<InstancesStandardizer, List<InstanceDescriptor>>>();
|
---|
| 164 | Action<InstanceExplorer> trainingBody = (kbExplorer) => {
|
---|
| 165 | var trainingData = GenerateData(instances, kbExplorer, parallel);
|
---|
| 166 | var standardizer = InstancesStandardizer.Create(trainingData);
|
---|
| 167 | standardizer.Apply(trainingData);
|
---|
| 168 | lock (sync) {
|
---|
| 169 | knowledgeBase.Add(kbExplorer, Tuple.Create(standardizer, trainingData));
|
---|
| 170 | }
|
---|
| 171 | };
|
---|
| 172 | if (parallel) {
|
---|
| 173 | Parallel.ForEach(trainingExplorers, trainingBody);
|
---|
| 174 | } else {
|
---|
| 175 | foreach (var kbExplorer in trainingExplorers) trainingBody(kbExplorer);
|
---|
[14690] | 176 | }
|
---|
[14691] | 177 |
|
---|
| 178 | var experimentBase = new Dictionary<InstanceExplorer, List<InstanceDescriptor>>();
|
---|
| 179 | Action<InstanceExplorer> testBody = (expExplorer) => {
|
---|
[14696] | 180 | var testData = GenerateData(instances, expExplorer, parallel);
|
---|
[14691] | 181 | lock (sync) {
|
---|
[14696] | 182 | experimentBase.Add(expExplorer, testData);
|
---|
[14691] | 183 | }
|
---|
| 184 | };
|
---|
| 185 | if (parallel) {
|
---|
| 186 | Parallel.ForEach(testExporers, testBody);
|
---|
| 187 | } else {
|
---|
| 188 | foreach (var expExplorer in testExporers) testBody(expExplorer);
|
---|
| 189 | }
|
---|
| 190 |
|
---|
| 191 | var data = from kb in knowledgeBase
|
---|
| 192 | from exp in experimentBase
|
---|
| 193 | select new { Training = kb, Test = exp };
|
---|
| 194 |
|
---|
| 195 | if (parallel) {
|
---|
| 196 | Parallel.ForEach(data, (point) => {
|
---|
| 197 | var normalizedTest = point.Test.Value.Select(x => new InstanceDescriptor(x)).ToList();
|
---|
| 198 | point.Training.Value.Item1.Apply(normalizedTest);
|
---|
| 199 | var result = Compare(point.Training.Value.Item2, normalizedTest);
|
---|
| 200 | lock (sync) {
|
---|
| 201 | Console.WriteLine("{0}\t{1}\t{2}\t{3:F2}\t{4}\t{5:F2}\t{6}",
|
---|
| 202 | point.Training.Key.Effort, point.Test.Key.Effort, result.ExactCount,
|
---|
| 203 | result.ExactAverageRank, result.ClsCount, result.ClsAverageRank, result.TotalCount);
|
---|
[14678] | 204 | }
|
---|
| 205 | });
|
---|
[14691] | 206 | } else {
|
---|
| 207 | foreach (var point in data) {
|
---|
| 208 | var normalizedTest = point.Test.Value.Select(x => new InstanceDescriptor(x)).ToList();
|
---|
| 209 | point.Training.Value.Item1.Apply(normalizedTest);
|
---|
| 210 | var result = Compare(point.Training.Value.Item2, normalizedTest);
|
---|
| 211 | Console.WriteLine("{0}\t{1}\t{2}\t{3:F2}\t{4}\t{5:F2}\t{6}",
|
---|
| 212 | point.Training.Key.Effort, point.Test.Key.Effort, result.ExactCount,
|
---|
| 213 | result.ExactAverageRank, result.ClsCount, result.ClsAverageRank, result.TotalCount);
|
---|
| 214 | }
|
---|
[14678] | 215 | }
|
---|
| 216 | }
|
---|
[14691] | 217 |
|
---|
| 218 | private class MatchResult {
|
---|
| 219 | public int ExactCount { get; set; }
|
---|
| 220 | public int ClsCount { get; set; }
|
---|
| 221 | public int TotalCount { get; set; }
|
---|
| 222 | public double ExactAverageRank { get; set; }
|
---|
| 223 | public double ClsAverageRank { get; set; }
|
---|
| 224 | }
|
---|
[14678] | 225 | }
|
---|
| 226 | }
|
---|