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 HeuristicLab.Analysis;
|
---|
23 | using HeuristicLab.Analysis.SelfOrganizingMaps;
|
---|
24 | using HeuristicLab.Collections;
|
---|
25 | using HeuristicLab.Common;
|
---|
26 | using HeuristicLab.Common.Resources;
|
---|
27 | using HeuristicLab.Core;
|
---|
28 | using HeuristicLab.Data;
|
---|
29 | using HeuristicLab.MainForm;
|
---|
30 | using HeuristicLab.Optimization;
|
---|
31 | using HeuristicLab.Persistence.Default.Xml;
|
---|
32 | using HeuristicLab.Random;
|
---|
33 | using System;
|
---|
34 | using System.Collections.Generic;
|
---|
35 | using System.Drawing;
|
---|
36 | using System.IO;
|
---|
37 | using System.Linq;
|
---|
38 | using System.Threading;
|
---|
39 | using System.Threading.Tasks;
|
---|
40 | using RunCreationClient = HeuristicLab.Clients.OKB.RunCreation.RunCreationClient;
|
---|
41 | using SingleObjectiveOKBProblem = HeuristicLab.Clients.OKB.RunCreation.SingleObjectiveOKBProblem;
|
---|
42 | using SingleObjectiveOKBSolution = HeuristicLab.Clients.OKB.RunCreation.SingleObjectiveOKBSolution;
|
---|
43 |
|
---|
44 | namespace HeuristicLab.OptimizationExpertSystem.Common {
|
---|
45 | [Item("Knowledge Center", "Currently in experimental phase, an expert system that makes algorithm suggestions based on fitness landscape analysis features and an optimization knowledge base.")]
|
---|
46 | [Creatable(CreatableAttribute.Categories.TestingAndAnalysis, Priority = 119)]
|
---|
47 | public sealed class KnowledgeCenter : IContent {
|
---|
48 | private bool SuppressEvents { get; set; }
|
---|
49 |
|
---|
50 | public string Filename { get; set; }
|
---|
51 |
|
---|
52 | public static new Image StaticItemImage {
|
---|
53 | get { return VSImageLibrary.Library; }
|
---|
54 | }
|
---|
55 |
|
---|
56 | private readonly IntValue maximumEvaluations;
|
---|
57 | public IntValue MaximumEvaluations {
|
---|
58 | get { return maximumEvaluations; }
|
---|
59 | }
|
---|
60 |
|
---|
61 | private readonly DoubleValue minimumTarget;
|
---|
62 | public DoubleValue MinimumTarget {
|
---|
63 | get { return minimumTarget; }
|
---|
64 | }
|
---|
65 |
|
---|
66 | private readonly RunCollection instanceRuns;
|
---|
67 | public RunCollection InstanceRuns {
|
---|
68 | get { return instanceRuns; }
|
---|
69 | }
|
---|
70 |
|
---|
71 | private readonly RunCollection seededRuns;
|
---|
72 | public RunCollection SeededRuns {
|
---|
73 | get { return seededRuns; }
|
---|
74 | }
|
---|
75 |
|
---|
76 | private readonly RunCollection knowledgeBase;
|
---|
77 | public RunCollection KnowledgeBase {
|
---|
78 | get { return knowledgeBase; }
|
---|
79 | }
|
---|
80 |
|
---|
81 | private readonly SingleObjectiveOKBProblem problem;
|
---|
82 | public SingleObjectiveOKBProblem Problem {
|
---|
83 | get { return problem; }
|
---|
84 | }
|
---|
85 |
|
---|
86 | private readonly ItemList<IAlgorithm> suggestedInstances;
|
---|
87 | private readonly ReadOnlyItemList<IAlgorithm> readOnlySuggestedInstances;
|
---|
88 | public ReadOnlyItemList<IAlgorithm> SuggestedInstances {
|
---|
89 | get { return readOnlySuggestedInstances; }
|
---|
90 | }
|
---|
91 |
|
---|
92 | private readonly RunCollection problemInstances;
|
---|
93 | public RunCollection ProblemInstances {
|
---|
94 | get { return problemInstances; }
|
---|
95 | }
|
---|
96 |
|
---|
97 | private readonly CheckedItemList<StringValue> problemCharacteristics;
|
---|
98 | private readonly ReadOnlyCheckedItemList<StringValue> readonlyProblemCharacteristics;
|
---|
99 | public ReadOnlyCheckedItemList<StringValue> ProblemCharacteristics {
|
---|
100 | get { return readonlyProblemCharacteristics; }
|
---|
101 | }
|
---|
102 |
|
---|
103 | private readonly EnumValue<ProblemInstanceProximityType> problemInstanceProximity;
|
---|
104 | public EnumValue<ProblemInstanceProximityType> ProblemInstanceProximity {
|
---|
105 | get { return problemInstanceProximity; }
|
---|
106 | }
|
---|
107 |
|
---|
108 | private readonly DoubleValue problemInstanceNeighborhoodFactor;
|
---|
109 | public DoubleValue ProblemInstanceNeighborhoodFactor {
|
---|
110 | get { return problemInstanceNeighborhoodFactor; }
|
---|
111 | }
|
---|
112 |
|
---|
113 | private readonly CheckedItemList<IScope> solutionSeedingPool;
|
---|
114 | public CheckedItemList<IScope> SolutionSeedingPool {
|
---|
115 | get { return solutionSeedingPool; }
|
---|
116 | }
|
---|
117 |
|
---|
118 | private readonly EnumValue<SeedingStrategyTypes> seedingStrategy;
|
---|
119 | public EnumValue<SeedingStrategyTypes> SeedingStrategy {
|
---|
120 | get { return seedingStrategy; }
|
---|
121 | }
|
---|
122 |
|
---|
123 | private BidirectionalLookup<long, IRun> algorithmId2RunMapping;
|
---|
124 | private BidirectionalDictionary<long, IAlgorithm> algorithmId2AlgorithmInstanceMapping;
|
---|
125 | private BidirectionalDictionary<long, IRun> problemId2ProblemInstanceMapping;
|
---|
126 |
|
---|
127 | private bool Maximization {
|
---|
128 | get { return Problem != null && Problem.ProblemId >= 0 && ((IValueParameter<BoolValue>)Problem.MaximizationParameter).Value.Value; }
|
---|
129 | }
|
---|
130 |
|
---|
131 | public KnowledgeCenter() {
|
---|
132 | maximumEvaluations = new IntValue(10000);
|
---|
133 | minimumTarget = new DoubleValue(1.05);
|
---|
134 | instanceRuns = new RunCollection();
|
---|
135 | seededRuns = new RunCollection();
|
---|
136 | knowledgeBase = new RunCollection();
|
---|
137 | suggestedInstances = new ItemList<IAlgorithm>();
|
---|
138 | readOnlySuggestedInstances = suggestedInstances.AsReadOnly();
|
---|
139 | problemInstances = new RunCollection();
|
---|
140 | problemCharacteristics = new CheckedItemList<StringValue>();
|
---|
141 | problemInstanceProximity = new EnumValue<ProblemInstanceProximityType>(ProblemInstanceProximityType.FeatureSpace);
|
---|
142 | problemInstanceNeighborhoodFactor = new DoubleValue(1);
|
---|
143 | readonlyProblemCharacteristics = problemCharacteristics.AsReadOnly();
|
---|
144 | problem = new SingleObjectiveOKBProblem();
|
---|
145 | algorithmId2RunMapping = new BidirectionalLookup<long, IRun>();
|
---|
146 | algorithmId2AlgorithmInstanceMapping = new BidirectionalDictionary<long, IAlgorithm>();
|
---|
147 | problemId2ProblemInstanceMapping = new BidirectionalDictionary<long, IRun>();
|
---|
148 | solutionSeedingPool = new CheckedItemList<IScope>();
|
---|
149 | seedingStrategy = new EnumValue<SeedingStrategyTypes>(SeedingStrategyTypes.NoSeeding);
|
---|
150 | RegisterEventHandlers();
|
---|
151 | }
|
---|
152 |
|
---|
153 | private void ProblemOnProblemChanged(object sender, EventArgs eventArgs) {
|
---|
154 | // TODO: Potentially, knowledge base has to be re-downloaded
|
---|
155 | }
|
---|
156 |
|
---|
157 | private void RegisterEventHandlers() {
|
---|
158 | maximumEvaluations.ValueChanged += MaximumEvaluationsOnValueChanged;
|
---|
159 | minimumTarget.ValueChanged += MinimumTargetOnValueChanged;
|
---|
160 | problem.ProblemChanged += ProblemOnProblemChanged;
|
---|
161 | problem.Solutions.ItemsAdded += ProblemSolutionsChanged;
|
---|
162 | problem.Solutions.ItemsReplaced += ProblemSolutionsChanged;
|
---|
163 | problem.Solutions.ItemsRemoved += ProblemSolutionsChanged;
|
---|
164 | problem.Solutions.CollectionReset += ProblemSolutionsChanged;
|
---|
165 | instanceRuns.CollectionReset += InformationChanged;
|
---|
166 | instanceRuns.ItemsAdded += InformationChanged;
|
---|
167 | instanceRuns.ItemsRemoved += InformationChanged;
|
---|
168 | instanceRuns.Reset += InformationChanged;
|
---|
169 | instanceRuns.UpdateOfRunsInProgressChanged += InformationChanged;
|
---|
170 | knowledgeBase.CollectionReset += InformationChanged;
|
---|
171 | knowledgeBase.ItemsAdded += InformationChanged;
|
---|
172 | knowledgeBase.ItemsRemoved += InformationChanged;
|
---|
173 | problemCharacteristics.ItemsAdded += CharacteristicChanged;
|
---|
174 | problemCharacteristics.ItemsReplaced += CharacteristicChanged;
|
---|
175 | problemCharacteristics.ItemsRemoved += CharacteristicChanged;
|
---|
176 | problemCharacteristics.CollectionReset += CharacteristicChanged;
|
---|
177 | problemCharacteristics.CheckedItemsChanged += CharacteristicChanged;
|
---|
178 | problemInstanceProximity.ValueChanged += ProblemInstanceProximityChanged;
|
---|
179 | problemInstanceNeighborhoodFactor.ValueChanged += ProblemInstanceProximityChanged;
|
---|
180 | }
|
---|
181 |
|
---|
182 | private void MaximumEvaluationsOnValueChanged(object sender, EventArgs eventArgs) {
|
---|
183 | UpdateSuggestions();
|
---|
184 | }
|
---|
185 |
|
---|
186 | private void MinimumTargetOnValueChanged(object sender, EventArgs e) {
|
---|
187 | UpdateSuggestions();
|
---|
188 | }
|
---|
189 |
|
---|
190 | private void ProblemSolutionsChanged(object sender, EventArgs e) {
|
---|
191 | foreach (var sol in Problem.Solutions.Select(x => x.Solution).OfType<IScope>()) {
|
---|
192 | if (!SolutionSeedingPool.Contains(sol))
|
---|
193 | SolutionSeedingPool.Add(sol, false);
|
---|
194 | }
|
---|
195 | }
|
---|
196 |
|
---|
197 | private void InformationChanged(object sender, EventArgs e) {
|
---|
198 | var runCollection = sender as RunCollection;
|
---|
199 | if (runCollection != null && runCollection.UpdateOfRunsInProgress) return;
|
---|
200 | UpdateSuggestions();
|
---|
201 | }
|
---|
202 |
|
---|
203 | private void CharacteristicChanged(object sender, EventArgs e) {
|
---|
204 | if (SuppressEvents) return;
|
---|
205 | UpdateInstanceProjection();
|
---|
206 | }
|
---|
207 |
|
---|
208 | private void ProblemInstanceProximityChanged(object sender, EventArgs e) {
|
---|
209 | UpdateSuggestions();
|
---|
210 | }
|
---|
211 |
|
---|
212 | public bool IsCurrentInstance(IRun run) {
|
---|
213 | if (!problemId2ProblemInstanceMapping.ContainsSecond(run)) return false;
|
---|
214 | return problemId2ProblemInstanceMapping.GetBySecond(run) == Problem.ProblemId;
|
---|
215 | }
|
---|
216 |
|
---|
217 | public void UpdateInstanceProjection() {
|
---|
218 | if (ProblemCharacteristics.Count == 0) return;
|
---|
219 |
|
---|
220 | var instances = GetProblemCharacteristics();
|
---|
221 |
|
---|
222 | var key2Idx = new BidirectionalDictionary<IRun, int>();
|
---|
223 | foreach (var kvp in instances.Select((k, i) => new { Index = i, Key = k.Key }))
|
---|
224 | key2Idx.Add(kvp.Key, kvp.Index);
|
---|
225 |
|
---|
226 | #region MDS
|
---|
227 | Func<double[], double[], double> euclid = (a, b) => Math.Sqrt(a.Zip(b, (x, y) => (x - y)).Sum(x => x * x));
|
---|
228 | var num = instances.Count;
|
---|
229 | var matrix = new DoubleMatrix(num, num);
|
---|
230 | for (var i = 0; i < num - 1; i++) {
|
---|
231 | for (var j = i + 1; j < num; j++) {
|
---|
232 | matrix[i, j] = matrix[j, i] = euclid(instances[key2Idx.GetBySecond(i)], instances[key2Idx.GetBySecond(j)]);
|
---|
233 | }
|
---|
234 | }
|
---|
235 |
|
---|
236 | var coords = MultidimensionalScaling.KruskalShepard(matrix);
|
---|
237 | #endregion
|
---|
238 | #region PCA
|
---|
239 | var ds = new double[instances.Count, ProblemCharacteristics.CheckedItems.Count()];
|
---|
240 | foreach (var instance in instances) {
|
---|
241 | var arr = instance.Value;
|
---|
242 | for (var feature = 0; feature < arr.Length; feature++)
|
---|
243 | ds[key2Idx.GetByFirst(instance.Key), feature] = arr[feature];
|
---|
244 | }
|
---|
245 |
|
---|
246 | int info;
|
---|
247 | double[] s2;
|
---|
248 | double[,] v;
|
---|
249 | alglib.pcabuildbasis(ds, ds.GetLength(0), ds.GetLength(1), out info, out s2, out v);
|
---|
250 | #endregion
|
---|
251 | #region SOM
|
---|
252 | var features = new DoubleMatrix(ProblemCharacteristics.CheckedItems.Count(), instances.Count);
|
---|
253 | foreach (var instance in instances) {
|
---|
254 | var arr = instance.Value;
|
---|
255 | for (var feature = 0; feature < arr.Length; feature++)
|
---|
256 | features[feature, key2Idx.GetByFirst(instance.Key)] = arr[feature];
|
---|
257 | }
|
---|
258 | var somCoords = SOM.Map(features, new MersenneTwister(42), somSize: 20, learningRadius: 20, iterations: 200, jittering: true);
|
---|
259 | #endregion
|
---|
260 |
|
---|
261 | ProblemInstances.UpdateOfRunsInProgress = true;
|
---|
262 | try {
|
---|
263 | foreach (var instance in ProblemInstances) {
|
---|
264 | double x = 0, y = 0;
|
---|
265 | for (var feature = 0; feature < ds.GetLength(1); feature++) {
|
---|
266 | x += ds[key2Idx.GetByFirst(instance), feature] * v[feature, 0];
|
---|
267 | y += ds[key2Idx.GetByFirst(instance), feature] * v[feature, 1];
|
---|
268 | }
|
---|
269 |
|
---|
270 | IItem item;
|
---|
271 | if (instance.Results.TryGetValue("Projection.PCA.X", out item)) {
|
---|
272 | ((DoubleValue)item).Value = x;
|
---|
273 | } else instance.Results.Add("Projection.PCA.X", new DoubleValue(x));
|
---|
274 | if (instance.Results.TryGetValue("Projection.PCA.Y", out item)) {
|
---|
275 | ((DoubleValue)item).Value = y;
|
---|
276 | } else instance.Results.Add("Projection.PCA.Y", new DoubleValue(y));
|
---|
277 |
|
---|
278 | if (instance.Results.TryGetValue("Projection.MDS.X", out item)) {
|
---|
279 | ((DoubleValue)item).Value = coords[key2Idx.GetByFirst(instance), 0];
|
---|
280 | } else instance.Results.Add("Projection.MDS.X", new DoubleValue(coords[key2Idx.GetByFirst(instance), 0]));
|
---|
281 | if (instance.Results.TryGetValue("Projection.MDS.Y", out item)) {
|
---|
282 | ((DoubleValue)item).Value = coords[key2Idx.GetByFirst(instance), 1];
|
---|
283 | } else instance.Results.Add("Projection.MDS.Y", new DoubleValue(coords[key2Idx.GetByFirst(instance), 1]));
|
---|
284 |
|
---|
285 | if (instance.Results.TryGetValue("Projection.SOM.X", out item)) {
|
---|
286 | ((DoubleValue)item).Value = somCoords[key2Idx.GetByFirst(instance), 0];
|
---|
287 | } else instance.Results.Add("Projection.SOM.X", new DoubleValue(somCoords[key2Idx.GetByFirst(instance), 0]));
|
---|
288 | if (instance.Results.TryGetValue("Projection.SOM.Y", out item)) {
|
---|
289 | ((DoubleValue)item).Value = somCoords[key2Idx.GetByFirst(instance), 1];
|
---|
290 | } else instance.Results.Add("Projection.SOM.Y", new DoubleValue(somCoords[key2Idx.GetByFirst(instance), 1]));
|
---|
291 | }
|
---|
292 | } finally { ProblemInstances.UpdateOfRunsInProgress = false; }
|
---|
293 | }
|
---|
294 |
|
---|
295 | private Dictionary<IRun, double[]> GetProblemCharacteristics() {
|
---|
296 | var instances = new Dictionary<IRun, double[]>();
|
---|
297 | var values = new List<double>[ProblemCharacteristics.CheckedItems.Count()];
|
---|
298 | foreach (var run in ProblemInstances) {
|
---|
299 | var f = 0;
|
---|
300 | instances[run] = new double[ProblemCharacteristics.CheckedItems.Count()];
|
---|
301 | foreach (var c in ProblemCharacteristics.CheckedItems.Select(x => x.Value.Value)) {
|
---|
302 | if (values[f] == null) values[f] = new List<double>(ProblemInstances.Count);
|
---|
303 | IItem item;
|
---|
304 | if (run.Results.TryGetValue(c, out item)) {
|
---|
305 | var val = (double)((dynamic)item).Value;
|
---|
306 | if (!double.IsNaN(val)) values[f].Add(val);
|
---|
307 | instances[run][f] = val;
|
---|
308 | } else instances[run][f] = double.NaN;
|
---|
309 | f++;
|
---|
310 | }
|
---|
311 | }
|
---|
312 | var median = values.Select(x => x.Count == 0 ? 0.0 : x.Median()).ToList();
|
---|
313 |
|
---|
314 | var allValues = instances.Values.Select(x => x.Select((f, i) => new {Idx = i, Val = double.IsNaN(f) ? median[i] : f}).ToList())
|
---|
315 | .SelectMany(x => x)
|
---|
316 | .GroupBy(x => x.Idx, x => x.Val)
|
---|
317 | .OrderBy(x => x.Key).ToList();
|
---|
318 | var avg = allValues.Select(x => x.Average()).ToList();
|
---|
319 | var stdev = allValues.Select(x => x.StandardDeviation()).ToList();
|
---|
320 |
|
---|
321 | // normalize characteristic values by transforming them to their z-score
|
---|
322 | foreach (var key in instances.Keys.ToList()) {
|
---|
323 | var arr = instances[key];
|
---|
324 | for (var i = 0; i < arr.Length; i++) {
|
---|
325 | if (double.IsNaN(arr[i])) arr[i] = median[i];
|
---|
326 | if (stdev[i] > 0) arr[i] = (arr[i] - avg[i]) / stdev[i];
|
---|
327 | }
|
---|
328 | }
|
---|
329 | return instances;
|
---|
330 | }
|
---|
331 |
|
---|
332 | private static readonly HashSet<string> InterestingValueNames = new HashSet<string>() {
|
---|
333 | "QualityPerEvaluations", "Problem Name", "Problem Type", "Algorithm Name", "Algorithm Type", "Maximization", "BestKnownQuality"
|
---|
334 | };
|
---|
335 |
|
---|
336 | public Task<ResultCollection> StartAlgorithmAsync(int index) {
|
---|
337 | return StartAlgorithmAsync(index, CancellationToken.None);
|
---|
338 | }
|
---|
339 |
|
---|
340 | public Task<ResultCollection> StartAlgorithmAsync(int index, CancellationToken cancellation) {
|
---|
341 | var selectedInstance = suggestedInstances[index];
|
---|
342 | var algorithmClone = (IAlgorithm)selectedInstance.Clone();
|
---|
343 | var problemClone = Problem.CloneProblem() as ISingleObjectiveHeuristicOptimizationProblem;
|
---|
344 | if (problemClone == null) throw new InvalidOperationException("Problem is not of type " + typeof(ISingleObjectiveHeuristicOptimizationProblem).FullName);
|
---|
345 | // TODO: It is assumed the problem instance by default is configured using no preexisting solution creator
|
---|
346 | var seedingStrategyLocal = SeedingStrategy.Value;
|
---|
347 | if (seedingStrategyLocal != SeedingStrategyTypes.NoSeeding) {
|
---|
348 | if (!SolutionSeedingPool.CheckedItems.Any()) throw new InvalidOperationException("There are no solutions selected for seeding.");
|
---|
349 | // TODO: It would be necessary to specify the solution creator somewhere (property and GUI)
|
---|
350 | var seedingCreator = problemClone.Operators.OfType<IPreexistingSolutionCreator>().FirstOrDefault();
|
---|
351 | if (seedingCreator == null) throw new InvalidOperationException("The problem does not contain a solution creator that allows seeding.");
|
---|
352 | seedingCreator.PreexistingSolutionsParameter.Value.Replace(SolutionSeedingPool.CheckedItems.Select(x => x.Value));
|
---|
353 | seedingCreator.SampleFromPreexistingParameter.Value.Value = seedingStrategyLocal == SeedingStrategyTypes.SeedBySampling;
|
---|
354 | // TODO: WHY!? WHY??!?
|
---|
355 | ((dynamic)problemClone.SolutionCreatorParameter).Value = (dynamic)seedingCreator;
|
---|
356 | }
|
---|
357 | algorithmClone.Problem = problemClone;
|
---|
358 | algorithmClone.Prepare(true);
|
---|
359 | IParameter stopParam;
|
---|
360 | var monitorStop = true;
|
---|
361 | if (algorithmClone.Parameters.TryGetValue("MaximumEvaluations", out stopParam)) {
|
---|
362 | var maxEvalParam = stopParam as IValueParameter<Data.IntValue>;
|
---|
363 | if (maxEvalParam != null) {
|
---|
364 | maxEvalParam.Value.Value = MaximumEvaluations.Value;
|
---|
365 | monitorStop = false;
|
---|
366 | }
|
---|
367 | }
|
---|
368 |
|
---|
369 | // TODO: The following can be simplified when we have async implementation patterns for our algorithms:
|
---|
370 | // TODO: The closures can be removed and replaced with private member methods
|
---|
371 | var waitHandle = new AutoResetEvent(false);
|
---|
372 |
|
---|
373 | #region EventHandler closures
|
---|
374 | EventHandler exeStateChanged = (sender, e) => {
|
---|
375 | if (algorithmClone.ExecutionState == ExecutionState.Stopped) {
|
---|
376 | lock (Problem.Solutions) {
|
---|
377 | foreach (var solution in algorithmClone.Results.Where(x => x.Name.ToLower().Contains("solution")).Select(x => x.Value).OfType<IScope>()) {
|
---|
378 | Problem.Solutions.Add(new SingleObjectiveOKBSolution(Problem.ProblemId) {
|
---|
379 | Quality = solution.Variables.ContainsKey(Problem.Problem.Evaluator.QualityParameter.ActualName) ? ((DoubleValue)solution.Variables[Problem.Problem.Evaluator.QualityParameter.ActualName].Value).Value : double.NaN,
|
---|
380 | Solution = (IItem)solution.Clone()
|
---|
381 | });
|
---|
382 | }
|
---|
383 | }
|
---|
384 | if (seedingStrategyLocal == SeedingStrategyTypes.NoSeeding) {
|
---|
385 | lock (InstanceRuns) {
|
---|
386 | InstanceRuns.Add(algorithmClone.Runs.Last());
|
---|
387 | }
|
---|
388 | } else {
|
---|
389 | lock (SeededRuns) {
|
---|
390 | SeededRuns.Add(algorithmClone.Runs.Last());
|
---|
391 | }
|
---|
392 | }
|
---|
393 | waitHandle.Set();
|
---|
394 | }
|
---|
395 | };
|
---|
396 |
|
---|
397 | EventHandler<EventArgs<Exception>> exceptionOccurred = (sender, e) => {
|
---|
398 | waitHandle.Set();
|
---|
399 | };
|
---|
400 |
|
---|
401 | EventHandler timeChanged = (sender, e) => {
|
---|
402 | IResult evalSolResult;
|
---|
403 | if (!algorithmClone.Results.TryGetValue("EvaluatedSolutions", out evalSolResult) || !(evalSolResult.Value is Data.IntValue)) return;
|
---|
404 | var evalSols = ((Data.IntValue)evalSolResult.Value).Value;
|
---|
405 | if (evalSols >= MaximumEvaluations.Value && algorithmClone.ExecutionState == ExecutionState.Started)
|
---|
406 | algorithmClone.Stop();
|
---|
407 | };
|
---|
408 | #endregion
|
---|
409 |
|
---|
410 | algorithmClone.ExecutionStateChanged += exeStateChanged;
|
---|
411 | algorithmClone.ExceptionOccurred += exceptionOccurred;
|
---|
412 | if (monitorStop) algorithmClone.ExecutionTimeChanged += timeChanged;
|
---|
413 |
|
---|
414 | return Task.Factory.StartNew(() => {
|
---|
415 | algorithmClone.Start();
|
---|
416 | OnAlgorithmInstanceStarted(algorithmClone);
|
---|
417 | var cancelRequested = false;
|
---|
418 | while (!waitHandle.WaitOne(200)) {
|
---|
419 | if (cancellation.IsCancellationRequested) {
|
---|
420 | cancelRequested = true;
|
---|
421 | break;
|
---|
422 | }
|
---|
423 | }
|
---|
424 | if (cancelRequested) {
|
---|
425 | try { algorithmClone.Stop(); } catch { } // ignore race condition if it is stopped in the meantime
|
---|
426 | waitHandle.WaitOne();
|
---|
427 | }
|
---|
428 | waitHandle.Dispose();
|
---|
429 | return algorithmClone.Results;
|
---|
430 | }, TaskCreationOptions.LongRunning);
|
---|
431 | }
|
---|
432 |
|
---|
433 | public ResultCollection StartAlgorithm(int index, CancellationToken cancellation) {
|
---|
434 | var task = StartAlgorithmAsync(index, cancellation);
|
---|
435 | task.Wait(cancellation);
|
---|
436 | return task.Result;
|
---|
437 | }
|
---|
438 |
|
---|
439 | public Task UpdateKnowledgeBaseAsync(IProgress progress = null) {
|
---|
440 | if (progress == null) progress = new Progress(string.Empty);
|
---|
441 | progress.Start("Updating Knowledge Base from OKB");
|
---|
442 | OnDownloadStarted(progress);
|
---|
443 | return Task.Factory.StartNew(() => { DoUpdateKnowledgeBase(progress); }, TaskCreationOptions.LongRunning);
|
---|
444 | }
|
---|
445 |
|
---|
446 | public void UpdateKnowledgeBase(IProgress progress = null) {
|
---|
447 | UpdateKnowledgeBaseAsync(progress).Wait();
|
---|
448 | }
|
---|
449 |
|
---|
450 | private void DoUpdateKnowledgeBase(IProgress progress) {
|
---|
451 | var queryClient = Clients.OKB.Query.QueryClient.Instance;
|
---|
452 | var adminClient = Clients.OKB.Administration.AdministrationClient.Instance;
|
---|
453 | try {
|
---|
454 | progress.Status = "Connecting to OKB...";
|
---|
455 | progress.ProgressValue = 0;
|
---|
456 | // FIXME: How to tell if refresh is necessary?
|
---|
457 | queryClient.Refresh();
|
---|
458 | progress.ProgressValue = 0.5;
|
---|
459 | progress.Status = "Downloading algorithm and problem instance information...";
|
---|
460 | // FIXME: How to tell if refresh is necessary?
|
---|
461 | adminClient.Refresh();
|
---|
462 |
|
---|
463 | var probInstance = adminClient.Problems.SingleOrDefault(x => x.Id == Problem.ProblemId);
|
---|
464 | if (probInstance == null) throw new InvalidOperationException("The chosen problem instance cannot be found in the OKB.");
|
---|
465 | var probClassId = probInstance.ProblemClassId;
|
---|
466 |
|
---|
467 | var problemClassFilter = (Clients.OKB.Query.StringComparisonAvailableValuesFilter)queryClient.Filters.Single(x => x.Label == "Problem Class Name");
|
---|
468 | problemClassFilter.Value = adminClient.ProblemClasses.Single(x => x.Id == probClassId).Name;
|
---|
469 |
|
---|
470 | problemId2ProblemInstanceMapping.Clear();
|
---|
471 | progress.Status = "Downloading problem instances...";
|
---|
472 | progress.ProgressValue = 0;
|
---|
473 | int[] p = { 0 };
|
---|
474 | ProblemInstances.UpdateOfRunsInProgress = true;
|
---|
475 | ProblemInstances.Clear();
|
---|
476 | var characteristics = new HashSet<string>();
|
---|
477 | var totalProblems = adminClient.Problems.Count(x => x.ProblemClassId == probClassId);
|
---|
478 | Parallel.ForEach(adminClient.Problems.Where(x => x.ProblemClassId == probClassId), (pInst) => {
|
---|
479 | var charas = new List<string>();
|
---|
480 | IRun probRun = null;
|
---|
481 | var data = Clients.OKB.Administration.AdministrationClient.GetProblemData(pInst.Id);
|
---|
482 | if (data != null) {
|
---|
483 | using (var stream = new MemoryStream(data)) {
|
---|
484 | try {
|
---|
485 | var prob = (IProblem)XmlParser.Deserialize<IContent>(stream);
|
---|
486 | probRun = new Run() { Name = prob.Name };
|
---|
487 | prob.CollectParameterValues(probRun.Parameters);
|
---|
488 | probRun.Parameters["Problem Name"] = new StringValue(prob.Name);
|
---|
489 | probRun.Parameters["Problem Type"] = new StringValue(prob.GetType().Name);
|
---|
490 | foreach (var v in RunCreationClient.Instance.GetCharacteristicValues(pInst.Id)) {
|
---|
491 | probRun.Results.Add("Characteristic." + v.Name, RunCreationClient.Instance.ConvertToItem(v));
|
---|
492 | charas.Add("Characteristic." + v.Name);
|
---|
493 | }
|
---|
494 | } catch { }
|
---|
495 | stream.Close();
|
---|
496 | }
|
---|
497 | if (probRun != null) {
|
---|
498 | lock (characteristics) {
|
---|
499 | problemId2ProblemInstanceMapping.Add(pInst.Id, probRun);
|
---|
500 | ProblemInstances.Add(probRun);
|
---|
501 | progress.Status = string.Format("Downloaded problem {0} (okb-id: {1})....", pInst.Name, pInst.Id);
|
---|
502 | p[0]++;
|
---|
503 | progress.ProgressValue = p[0] / (double)totalProblems;
|
---|
504 | foreach (var c in charas) characteristics.Add(c);
|
---|
505 | }
|
---|
506 | }
|
---|
507 | }
|
---|
508 | });
|
---|
509 |
|
---|
510 | algorithmId2AlgorithmInstanceMapping.Clear();
|
---|
511 | progress.Status = "Downloading algorithm instances...";
|
---|
512 | progress.ProgressValue = 0;
|
---|
513 | p[0] = 0;
|
---|
514 | Parallel.ForEach(adminClient.Algorithms, (algInst) => {
|
---|
515 | IAlgorithm alg = null;
|
---|
516 | var data = Clients.OKB.Administration.AdministrationClient.GetAlgorithmData(algInst.Id);
|
---|
517 | if (data != null) {
|
---|
518 | using (var stream = new MemoryStream(data)) {
|
---|
519 | try {
|
---|
520 | alg = (IAlgorithm)XmlParser.Deserialize<IContent>(stream);
|
---|
521 | } catch { }
|
---|
522 | stream.Close();
|
---|
523 | }
|
---|
524 | if (alg != null) {
|
---|
525 | lock (algorithmId2AlgorithmInstanceMapping) {
|
---|
526 | algorithmId2AlgorithmInstanceMapping.Add(algInst.Id, alg);
|
---|
527 | progress.Status = string.Format("Downloaded algorithm {0} (okb-id: {1})...", algInst.Name, algInst.Id);
|
---|
528 | p[0]++;
|
---|
529 | progress.ProgressValue = p[0] / (double)adminClient.Algorithms.Count;
|
---|
530 | }
|
---|
531 | }
|
---|
532 | }
|
---|
533 | });
|
---|
534 |
|
---|
535 | var interestingValues = queryClient.ValueNames.Where(x => InterestingValueNames.Contains(x.Name)).ToList();
|
---|
536 |
|
---|
537 | progress.Status = "Downloading runs...";
|
---|
538 | progress.ProgressValue = 0;
|
---|
539 | p[0] = 0;
|
---|
540 | var count = queryClient.GetNumberOfRuns(problemClassFilter);
|
---|
541 | if (count == 0) return;
|
---|
542 |
|
---|
543 | var runList = new List<IRun>();
|
---|
544 | var runIds = queryClient.GetRunIds(problemClassFilter).ToList();
|
---|
545 | var batches = runIds.Select((v, i) => new { Idx = i, Val = v }).GroupBy(x => x.Idx / 500, x => x.Val);
|
---|
546 | Parallel.ForEach(batches.Select(x => x.ToList()), (batch) => {
|
---|
547 | var okbRuns = queryClient.GetRunsWithValues(batch, true, interestingValues);
|
---|
548 | var hlRuns = okbRuns.AsParallel().Select(x => new { AlgorithmId = x.Algorithm.Id, Run = queryClient.ConvertToOptimizationRun(x) }).ToList();
|
---|
549 | lock (runList) {
|
---|
550 | foreach (var r in hlRuns) {
|
---|
551 | algorithmId2RunMapping.Add(r.AlgorithmId, r.Run);
|
---|
552 | runList.Add(r.Run);
|
---|
553 | }
|
---|
554 | progress.Status = string.Format("Downloaded runs {0} to {1} of {2}...", p[0], p[0] + batch.Count, count);
|
---|
555 | p[0] += batch.Count;
|
---|
556 | progress.ProgressValue = p[0] / (double)count;
|
---|
557 | }
|
---|
558 | });
|
---|
559 |
|
---|
560 | progress.Status = "Finishing...";
|
---|
561 | var algInstRunDict = runList.Where(x => x.Parameters.ContainsKey("Problem Name") && x.Parameters["Problem Name"] is StringValue)
|
---|
562 | .GroupBy(x => ((StringValue)x.Parameters["Problem Name"]).Value)
|
---|
563 | .ToDictionary(x => x.Key, x => x.GroupBy(y => ((StringValue)y.Parameters["Algorithm Name"]).Value)
|
---|
564 | .ToDictionary(y => y.Key, y => y.ToList()));
|
---|
565 |
|
---|
566 | foreach (var instance in ProblemInstances) {
|
---|
567 | IItem probNameParam;
|
---|
568 | if (!instance.Parameters.TryGetValue("Problem Name", out probNameParam)) continue;
|
---|
569 |
|
---|
570 | var probInstanceName = ((StringValue)probNameParam).Value;
|
---|
571 | var maximization = ((BoolValue)instance.Parameters["Maximization"]).Value;
|
---|
572 |
|
---|
573 | IItem bkParam;
|
---|
574 | if (!instance.Parameters.TryGetValue("BestKnownQuality", out bkParam) || !(bkParam is DoubleValue)) {
|
---|
575 | Dictionary<string, List<IRun>> algRuns;
|
---|
576 | if (!algInstRunDict.TryGetValue(probInstanceName, out algRuns)) continue;
|
---|
577 | var list = algRuns.SelectMany(x => x.Value)
|
---|
578 | .Where(x => x.Results.ContainsKey("QualityPerEvaluations"))
|
---|
579 | .Select(x => ((IndexedDataTable<double>)x.Results["QualityPerEvaluations"]).Rows.First().Values.Last().Item2);
|
---|
580 | bkParam = new DoubleValue(maximization ? list.Max() : list.Min());
|
---|
581 | instance.Parameters["BestKnownQuality"] = bkParam;
|
---|
582 | } else bkParam = instance.Parameters["BestKnownQuality"];
|
---|
583 |
|
---|
584 | var bkQuality = ((DoubleValue)bkParam).Value;
|
---|
585 |
|
---|
586 | if (!algInstRunDict.ContainsKey(probInstanceName)) continue;
|
---|
587 | // TODO: Things needs to be configurable here (table name, targets)
|
---|
588 | foreach (var target in new[] { 1, 1.01, 1.05, 1.1, 1.2, 1.5 }) {
|
---|
589 | var indexMap = new BidirectionalDictionary<string, int>();
|
---|
590 | var dict = new Dictionary<string, double>();
|
---|
591 | var idx = 0;
|
---|
592 | foreach (var kvp in algInstRunDict[probInstanceName]) {
|
---|
593 | var result = ExpectedRuntimeHelper.CalculateErt(kvp.Value, "QualityPerEvaluations", bkQuality * target, maximization);
|
---|
594 | indexMap.Add(kvp.Key, idx);
|
---|
595 | dict[kvp.Key] = !double.IsNaN(result.ExpectedRuntime) ? result.ExpectedRuntime : int.MaxValue;
|
---|
596 | idx++;
|
---|
597 | }
|
---|
598 | var points = dict.OrderBy(x => indexMap.GetByFirst(x.Key)).Select(x => x.Value > 0 ? Math.Log10(x.Value) : 0).ToArray();
|
---|
599 | int[] clusters;
|
---|
600 | Ckmeans1dClusters(points, 5, out clusters);
|
---|
601 | var ranks = clusters.Select((c, i) => new { Cluster = c, Perf = dict[indexMap.GetBySecond(i)] })
|
---|
602 | .GroupBy(x => x.Cluster, x => x.Perf)
|
---|
603 | .Select(x => new { Cluster = x.Key, AvgPerf = x.Average() })
|
---|
604 | .OrderBy(x => x.AvgPerf)
|
---|
605 | .Select((c, i) => new { Cluster = c.Cluster, Rank = i })
|
---|
606 | .ToDictionary(x => x.Cluster, x => x.Rank);
|
---|
607 | for (var i = 0; i < clusters.Length; i++) {
|
---|
608 | var resultName = "Rank." + indexMap.GetBySecond(i) + "@" + ((target - 1) * 100) + "%";
|
---|
609 | instance.Results[resultName] = new IntValue(dict[indexMap.GetBySecond(i)] < int.MaxValue ? ranks[clusters[i]] : 6);
|
---|
610 | }
|
---|
611 | }
|
---|
612 | }
|
---|
613 | try {
|
---|
614 | SuppressEvents = true;
|
---|
615 | problemCharacteristics.Replace(characteristics.Select(x => new StringValue(x)));
|
---|
616 | foreach (var pc in problemCharacteristics.Where(x => !x.Value.StartsWith("Characteristic.")))
|
---|
617 | problemCharacteristics.SetItemCheckedState(pc, false);
|
---|
618 | } finally { SuppressEvents = false; }
|
---|
619 | try {
|
---|
620 | KnowledgeBase.UpdateOfRunsInProgress = true;
|
---|
621 | KnowledgeBase.Clear();
|
---|
622 | KnowledgeBase.AddRange(runList);
|
---|
623 | } finally { KnowledgeBase.UpdateOfRunsInProgress = false; }
|
---|
624 | } finally { progress.Finish(); ProblemInstances.UpdateOfRunsInProgress = false; }
|
---|
625 | UpdateInstanceProjection();
|
---|
626 | UpdateSuggestions();
|
---|
627 | }
|
---|
628 |
|
---|
629 | private void UpdateSuggestions() {
|
---|
630 | if (Problem == null) return;
|
---|
631 | var piDistances = GetProblemDistances();
|
---|
632 | var maxDistance = piDistances.Max();
|
---|
633 | // Weighted combination of algorithm performances using distance-based exponentially falling weights
|
---|
634 | // Scale distances by maxDistance
|
---|
635 | // Parameter to influence dampening factor
|
---|
636 | // Algorithm performances are given in expected run time
|
---|
637 | // Using convergence graphs up to maximum evaluations effort
|
---|
638 | // Care has to be taken if not every algorithm has been executed on every problem instance
|
---|
639 | var instances = new SortedList<double, IAlgorithm>();
|
---|
640 | foreach (var relevantRuns in knowledgeBase.GroupBy(x => algorithmId2RunMapping.GetBySecond(x).Single())) {
|
---|
641 | var algorithm = algorithmId2AlgorithmInstanceMapping.GetByFirst(relevantRuns.Key);
|
---|
642 | var avgQuality = 0.0;
|
---|
643 | var counter = 0;
|
---|
644 | foreach (var problemRuns in relevantRuns.GroupBy(x => ((StringValue)x.Parameters["Problem Name"]).Value)) {
|
---|
645 | var probInstance = ProblemInstances.SingleOrDefault(x => x.Name == problemRuns.Key);
|
---|
646 | if (probInstance == null) continue;
|
---|
647 | var bkQuality = ((DoubleValue)probInstance.Parameters["BestKnownQuality"]).Value;
|
---|
648 | foreach (var run in problemRuns) {
|
---|
649 | var performanceGraph = ((IndexedDataTable<double>)run.Results["QualityPerEvaluations"]);
|
---|
650 | try {
|
---|
651 | avgQuality += performanceGraph.Rows.First().Values.TakeWhile(x => x.Item1 < MaximumEvaluations.Value).Last().Item2 / bkQuality;
|
---|
652 | counter++;
|
---|
653 | } catch {
|
---|
654 | continue;
|
---|
655 | }
|
---|
656 | }
|
---|
657 | }
|
---|
658 | avgQuality /= counter;
|
---|
659 | instances.Add(avgQuality, algorithm);
|
---|
660 | }
|
---|
661 |
|
---|
662 | var instanceLadder = instances.Select(x => (IAlgorithm)x.Value.Clone()).ToList();
|
---|
663 | if (Maximization) instanceLadder.Reverse();
|
---|
664 | suggestedInstances.Replace(instanceLadder);
|
---|
665 | }
|
---|
666 |
|
---|
667 | private DoubleMatrix GetProblemDistances() {
|
---|
668 | var matrix = new DoubleMatrix(ProblemInstances.Count, ProblemInstances.Count);
|
---|
669 | switch (ProblemInstanceProximity.Value) {
|
---|
670 | case ProblemInstanceProximityType.MDS:
|
---|
671 | case ProblemInstanceProximityType.PCA:
|
---|
672 | case ProblemInstanceProximityType.SOM:
|
---|
673 | int i = 0, j = 0;
|
---|
674 | foreach (var a in ProblemInstances) {
|
---|
675 | double xa, ya;
|
---|
676 | GetProjectionCoordinates(a, out xa, out ya);
|
---|
677 | j = 0;
|
---|
678 | foreach (var b in ProblemInstances) {
|
---|
679 | double xb, yb;
|
---|
680 | GetProjectionCoordinates(b, out xb, out yb);
|
---|
681 | matrix[i, j] = Math.Sqrt((xa - xb) * (xa - xb) + (ya - yb) * (ya - yb));
|
---|
682 | j++;
|
---|
683 | }
|
---|
684 | i++;
|
---|
685 | }
|
---|
686 | break;
|
---|
687 | case ProblemInstanceProximityType.FeatureSpace:
|
---|
688 | int k = 0, l = 0;
|
---|
689 | var features = GetProblemCharacteristics();
|
---|
690 | foreach (var a in ProblemInstances) {
|
---|
691 | l = 0;
|
---|
692 | var fa = features[a];
|
---|
693 | foreach (var b in ProblemInstances) {
|
---|
694 | var sum = features[b].Select((t, f) => (fa[f] - t) * (fa[f] - t)).Sum();
|
---|
695 | matrix[k, l] = Math.Sqrt(sum);
|
---|
696 | l++;
|
---|
697 | }
|
---|
698 | k++;
|
---|
699 | }
|
---|
700 | break;
|
---|
701 | default: throw new InvalidOperationException(string.Format("Unkonwn proximity type {0}", ProblemInstanceProximity.Value));
|
---|
702 | }
|
---|
703 | return matrix;
|
---|
704 | }
|
---|
705 |
|
---|
706 | private void GetProjectionCoordinates(IRun problemInstance, out double x, out double y) {
|
---|
707 | x = ((DoubleValue)problemInstance.Results["Projection." + ProblemInstanceProximity.Value + ".X"]).Value;
|
---|
708 | y = ((DoubleValue)problemInstance.Results["Projection." + ProblemInstanceProximity.Value + ".Y"]).Value;
|
---|
709 | }
|
---|
710 |
|
---|
711 | public event EventHandler<EventArgs<IProgress>> DownloadStarted;
|
---|
712 | private void OnDownloadStarted(IProgress progress) {
|
---|
713 | var handler = DownloadStarted;
|
---|
714 | if (handler != null) handler(this, new EventArgs<IProgress>(progress));
|
---|
715 | }
|
---|
716 |
|
---|
717 | public event EventHandler<EventArgs<IAlgorithm>> AlgorithmInstanceStarted;
|
---|
718 | private void OnAlgorithmInstanceStarted(IAlgorithm instance) {
|
---|
719 | var handler = AlgorithmInstanceStarted;
|
---|
720 | if (handler != null) handler(this, new EventArgs<IAlgorithm>(instance));
|
---|
721 | }
|
---|
722 |
|
---|
723 | // implement further classes and methods
|
---|
724 | private static SortedList<double, int> Ckmeans1dClusters(double[] estimations, int k, out int[] clusterValues) {
|
---|
725 | int nPoints = estimations.Length;
|
---|
726 | var distinct = estimations.Distinct().OrderBy(x => x).ToArray();
|
---|
727 | var max = distinct.Max();
|
---|
728 | if (distinct.Length <= k) {
|
---|
729 | var dict = distinct.Select((v, i) => new { Index = i, Value = v }).ToDictionary(x => x.Value, y => y.Index);
|
---|
730 | for (int i = distinct.Length; i < k; i++)
|
---|
731 | dict.Add(max + i - distinct.Length + 1, i);
|
---|
732 |
|
---|
733 | clusterValues = new int[nPoints];
|
---|
734 | for (int i = 0; i < nPoints; i++)
|
---|
735 | if (!dict.ContainsKey(estimations[i])) clusterValues[i] = 0;
|
---|
736 | else clusterValues[i] = dict[estimations[i]];
|
---|
737 |
|
---|
738 | return new SortedList<double, int>(dict);
|
---|
739 | }
|
---|
740 |
|
---|
741 | var n = distinct.Length;
|
---|
742 | var D = new double[n, k];
|
---|
743 | var B = new int[n, k];
|
---|
744 |
|
---|
745 | for (int m = 0; m < k; m++) {
|
---|
746 | for (int j = m; j <= n - k + m; j++) {
|
---|
747 | if (m == 0)
|
---|
748 | D[j, m] = SumOfSquaredDistances(distinct, 0, j + 1);
|
---|
749 | else {
|
---|
750 | var minD = double.MaxValue;
|
---|
751 | var minI = 0;
|
---|
752 | for (int i = 1; i <= j; i++) {
|
---|
753 | var d = D[i - 1, m - 1] + SumOfSquaredDistances(distinct, i, j + 1);
|
---|
754 | if (d < minD) {
|
---|
755 | minD = d;
|
---|
756 | minI = i;
|
---|
757 | }
|
---|
758 | }
|
---|
759 | D[j, m] = minD;
|
---|
760 | B[j, m] = minI;
|
---|
761 | }
|
---|
762 | }
|
---|
763 | }
|
---|
764 |
|
---|
765 | var centers = new SortedList<double, int>();
|
---|
766 | var upper = B[n - 1, k - 1];
|
---|
767 | var c = Mean(distinct, upper, n);
|
---|
768 | centers.Add(c, k - 1);
|
---|
769 | for (int i = k - 2; i >= 0; i--) {
|
---|
770 | var lower = B[upper - 1, i];
|
---|
771 | var c2 = Mean(distinct, lower, upper);
|
---|
772 | centers.Add(c2, i);
|
---|
773 | upper = lower;
|
---|
774 | }
|
---|
775 |
|
---|
776 | clusterValues = new int[nPoints];
|
---|
777 | for (int i = 0; i < estimations.Length; i++) {
|
---|
778 | clusterValues[i] = centers.MinItems(x => Math.Abs(estimations[i] - x.Key)).First().Value;
|
---|
779 | }
|
---|
780 |
|
---|
781 | return centers;
|
---|
782 | }
|
---|
783 |
|
---|
784 | private static double SumOfSquaredDistances(double[] x, int start, int end) {
|
---|
785 | if (start == end) throw new InvalidOperationException();
|
---|
786 | if (start + 1 == end) return 0.0;
|
---|
787 | double mean = 0.0;
|
---|
788 | for (int i = start; i < end; i++) {
|
---|
789 | mean += x[i];
|
---|
790 | }
|
---|
791 | mean /= (end - start);
|
---|
792 | var sum = 0.0;
|
---|
793 | for (int i = start; i < end; i++) {
|
---|
794 | sum += (x[i] - mean) * (x[i] - mean);
|
---|
795 | }
|
---|
796 | return sum;
|
---|
797 | }
|
---|
798 |
|
---|
799 | private static double Mean(double[] x, int start, int end) {
|
---|
800 | if (start == end) throw new InvalidOperationException();
|
---|
801 | double mean = 0.0;
|
---|
802 | for (int i = start; i < end; i++) {
|
---|
803 | mean += x[i];
|
---|
804 | }
|
---|
805 | mean /= (end - start);
|
---|
806 | return mean;
|
---|
807 | }
|
---|
808 | }
|
---|
809 | }
|
---|