| 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;
|
|---|
| 23 | using System.Collections.Generic;
|
|---|
| 24 | using System.Drawing;
|
|---|
| 25 | using System.Linq;
|
|---|
| 26 | using System.Threading;
|
|---|
| 27 | using HeuristicLab.Analysis;
|
|---|
| 28 | using HeuristicLab.Common;
|
|---|
| 29 | using HeuristicLab.Core;
|
|---|
| 30 | using HeuristicLab.Data;
|
|---|
| 31 | using HeuristicLab.Optimization;
|
|---|
| 32 | using HeuristicLab.Parameters;
|
|---|
| 33 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
|---|
| 34 | using HeuristicLab.PluginInfrastructure;
|
|---|
| 35 | using HeuristicLab.Problems.DataAnalysis;
|
|---|
| 36 | using HeuristicLab.Random;
|
|---|
| 37 |
|
|---|
| 38 | namespace HeuristicLab.Algorithms.DataAnalysis {
|
|---|
| 39 | /// <summary>
|
|---|
| 40 | /// t-distributed stochastic neighbourhood embedding (tSNE) projects the data in a low dimensional
|
|---|
| 41 | /// space to allow visual cluster identification.
|
|---|
| 42 | /// </summary>
|
|---|
| 43 | [Item("tSNE", "t-distributed stochastic neighbourhood embedding projects the data in a low " +
|
|---|
| 44 | "dimensional space to allow visual cluster identification. Implemented similar to: https://lvdmaaten.github.io/tsne/#implementations (Barnes-Hut t-SNE). Described in : https://lvdmaaten.github.io/publications/papers/JMLR_2014.pdf")]
|
|---|
| 45 | [Creatable(CreatableAttribute.Categories.DataAnalysis, Priority = 100)]
|
|---|
| 46 | [StorableClass]
|
|---|
| 47 | public sealed class TSNEAlgorithm : BasicAlgorithm {
|
|---|
| 48 | public override bool SupportsPause {
|
|---|
| 49 | get { return true; }
|
|---|
| 50 | }
|
|---|
| 51 | public override Type ProblemType {
|
|---|
| 52 | get { return typeof(IDataAnalysisProblem); }
|
|---|
| 53 | }
|
|---|
| 54 | public new IDataAnalysisProblem Problem {
|
|---|
| 55 | get { return (IDataAnalysisProblem) base.Problem; }
|
|---|
| 56 | set { base.Problem = value; }
|
|---|
| 57 | }
|
|---|
| 58 |
|
|---|
| 59 | #region parameter names
|
|---|
| 60 | private const string DistanceFunctionParameterName = "DistanceFunction";
|
|---|
| 61 | private const string PerplexityParameterName = "Perplexity";
|
|---|
| 62 | private const string ThetaParameterName = "Theta";
|
|---|
| 63 | private const string NewDimensionsParameterName = "Dimensions";
|
|---|
| 64 | private const string MaxIterationsParameterName = "MaxIterations";
|
|---|
| 65 | private const string StopLyingIterationParameterName = "StopLyingIteration";
|
|---|
| 66 | private const string MomentumSwitchIterationParameterName = "MomentumSwitchIteration";
|
|---|
| 67 | private const string InitialMomentumParameterName = "InitialMomentum";
|
|---|
| 68 | private const string FinalMomentumParameterName = "FinalMomentum";
|
|---|
| 69 | private const string EtaParameterName = "Eta";
|
|---|
| 70 | private const string SetSeedRandomlyParameterName = "SetSeedRandomly";
|
|---|
| 71 | private const string SeedParameterName = "Seed";
|
|---|
| 72 | private const string ClassesNameParameterName = "ClassesName";
|
|---|
| 73 | private const string NormalizationParameterName = "Normalization";
|
|---|
| 74 | private const string RandomInitializationParameterName = "RandomInitialization";
|
|---|
| 75 | private const string UpdateIntervalParameterName = "UpdateInterval";
|
|---|
| 76 | #endregion
|
|---|
| 77 |
|
|---|
| 78 | #region result names
|
|---|
| 79 | private const string IterationResultName = "Iteration";
|
|---|
| 80 | private const string ErrorResultName = "Error";
|
|---|
| 81 | private const string ErrorPlotResultName = "Error plot";
|
|---|
| 82 | private const string ScatterPlotResultName = "Scatterplot";
|
|---|
| 83 | private const string DataResultName = "Projected data";
|
|---|
| 84 | #endregion
|
|---|
| 85 |
|
|---|
| 86 | #region parameter properties
|
|---|
| 87 | public IFixedValueParameter<DoubleValue> PerplexityParameter {
|
|---|
| 88 | get { return Parameters[PerplexityParameterName] as IFixedValueParameter<DoubleValue>; }
|
|---|
| 89 | }
|
|---|
| 90 | public IFixedValueParameter<PercentValue> ThetaParameter {
|
|---|
| 91 | get { return Parameters[ThetaParameterName] as IFixedValueParameter<PercentValue>; }
|
|---|
| 92 | }
|
|---|
| 93 | public IFixedValueParameter<IntValue> NewDimensionsParameter {
|
|---|
| 94 | get { return Parameters[NewDimensionsParameterName] as IFixedValueParameter<IntValue>; }
|
|---|
| 95 | }
|
|---|
| 96 | public IConstrainedValueParameter<IDistance<double[]>> DistanceFunctionParameter {
|
|---|
| 97 | get { return Parameters[DistanceFunctionParameterName] as IConstrainedValueParameter<IDistance<double[]>>; }
|
|---|
| 98 | }
|
|---|
| 99 | public IFixedValueParameter<IntValue> MaxIterationsParameter {
|
|---|
| 100 | get { return Parameters[MaxIterationsParameterName] as IFixedValueParameter<IntValue>; }
|
|---|
| 101 | }
|
|---|
| 102 | public IFixedValueParameter<IntValue> StopLyingIterationParameter {
|
|---|
| 103 | get { return Parameters[StopLyingIterationParameterName] as IFixedValueParameter<IntValue>; }
|
|---|
| 104 | }
|
|---|
| 105 | public IFixedValueParameter<IntValue> MomentumSwitchIterationParameter {
|
|---|
| 106 | get { return Parameters[MomentumSwitchIterationParameterName] as IFixedValueParameter<IntValue>; }
|
|---|
| 107 | }
|
|---|
| 108 | public IFixedValueParameter<DoubleValue> InitialMomentumParameter {
|
|---|
| 109 | get { return Parameters[InitialMomentumParameterName] as IFixedValueParameter<DoubleValue>; }
|
|---|
| 110 | }
|
|---|
| 111 | public IFixedValueParameter<DoubleValue> FinalMomentumParameter {
|
|---|
| 112 | get { return Parameters[FinalMomentumParameterName] as IFixedValueParameter<DoubleValue>; }
|
|---|
| 113 | }
|
|---|
| 114 | public IFixedValueParameter<DoubleValue> EtaParameter {
|
|---|
| 115 | get { return Parameters[EtaParameterName] as IFixedValueParameter<DoubleValue>; }
|
|---|
| 116 | }
|
|---|
| 117 | public IFixedValueParameter<BoolValue> SetSeedRandomlyParameter {
|
|---|
| 118 | get { return Parameters[SetSeedRandomlyParameterName] as IFixedValueParameter<BoolValue>; }
|
|---|
| 119 | }
|
|---|
| 120 | public IFixedValueParameter<IntValue> SeedParameter {
|
|---|
| 121 | get { return Parameters[SeedParameterName] as IFixedValueParameter<IntValue>; }
|
|---|
| 122 | }
|
|---|
| 123 | public IConstrainedValueParameter<StringValue> ClassesNameParameter {
|
|---|
| 124 | get { return Parameters[ClassesNameParameterName] as IConstrainedValueParameter<StringValue>; }
|
|---|
| 125 | }
|
|---|
| 126 | public IFixedValueParameter<BoolValue> NormalizationParameter {
|
|---|
| 127 | get { return Parameters[NormalizationParameterName] as IFixedValueParameter<BoolValue>; }
|
|---|
| 128 | }
|
|---|
| 129 | public IFixedValueParameter<BoolValue> RandomInitializationParameter {
|
|---|
| 130 | get { return Parameters[RandomInitializationParameterName] as IFixedValueParameter<BoolValue>; }
|
|---|
| 131 | }
|
|---|
| 132 | public IFixedValueParameter<IntValue> UpdateIntervalParameter {
|
|---|
| 133 | get { return Parameters[UpdateIntervalParameterName] as IFixedValueParameter<IntValue>; }
|
|---|
| 134 | }
|
|---|
| 135 | #endregion
|
|---|
| 136 |
|
|---|
| 137 | #region Properties
|
|---|
| 138 | public IDistance<double[]> DistanceFunction {
|
|---|
| 139 | get { return DistanceFunctionParameter.Value; }
|
|---|
| 140 | }
|
|---|
| 141 | public double Perplexity {
|
|---|
| 142 | get { return PerplexityParameter.Value.Value; }
|
|---|
| 143 | set { PerplexityParameter.Value.Value = value; }
|
|---|
| 144 | }
|
|---|
| 145 | public double Theta {
|
|---|
| 146 | get { return ThetaParameter.Value.Value; }
|
|---|
| 147 | set { ThetaParameter.Value.Value = value; }
|
|---|
| 148 | }
|
|---|
| 149 | public int NewDimensions {
|
|---|
| 150 | get { return NewDimensionsParameter.Value.Value; }
|
|---|
| 151 | set { NewDimensionsParameter.Value.Value = value; }
|
|---|
| 152 | }
|
|---|
| 153 | public int MaxIterations {
|
|---|
| 154 | get { return MaxIterationsParameter.Value.Value; }
|
|---|
| 155 | set { MaxIterationsParameter.Value.Value = value; }
|
|---|
| 156 | }
|
|---|
| 157 | public int StopLyingIteration {
|
|---|
| 158 | get { return StopLyingIterationParameter.Value.Value; }
|
|---|
| 159 | set { StopLyingIterationParameter.Value.Value = value; }
|
|---|
| 160 | }
|
|---|
| 161 | public int MomentumSwitchIteration {
|
|---|
| 162 | get { return MomentumSwitchIterationParameter.Value.Value; }
|
|---|
| 163 | set { MomentumSwitchIterationParameter.Value.Value = value; }
|
|---|
| 164 | }
|
|---|
| 165 | public double InitialMomentum {
|
|---|
| 166 | get { return InitialMomentumParameter.Value.Value; }
|
|---|
| 167 | set { InitialMomentumParameter.Value.Value = value; }
|
|---|
| 168 | }
|
|---|
| 169 | public double FinalMomentum {
|
|---|
| 170 | get { return FinalMomentumParameter.Value.Value; }
|
|---|
| 171 | set { FinalMomentumParameter.Value.Value = value; }
|
|---|
| 172 | }
|
|---|
| 173 | public double Eta {
|
|---|
| 174 | get { return EtaParameter.Value.Value; }
|
|---|
| 175 | set { EtaParameter.Value.Value = value; }
|
|---|
| 176 | }
|
|---|
| 177 | public bool SetSeedRandomly {
|
|---|
| 178 | get { return SetSeedRandomlyParameter.Value.Value; }
|
|---|
| 179 | set { SetSeedRandomlyParameter.Value.Value = value; }
|
|---|
| 180 | }
|
|---|
| 181 | public int Seed {
|
|---|
| 182 | get { return SeedParameter.Value.Value; }
|
|---|
| 183 | set { SeedParameter.Value.Value = value; }
|
|---|
| 184 | }
|
|---|
| 185 | public string ClassesName {
|
|---|
| 186 | get { return ClassesNameParameter.Value != null ? ClassesNameParameter.Value.Value : null; }
|
|---|
| 187 | set { ClassesNameParameter.Value.Value = value; }
|
|---|
| 188 | }
|
|---|
| 189 | public bool Normalization {
|
|---|
| 190 | get { return NormalizationParameter.Value.Value; }
|
|---|
| 191 | set { NormalizationParameter.Value.Value = value; }
|
|---|
| 192 | }
|
|---|
| 193 | public bool RandomInitialization {
|
|---|
| 194 | get { return RandomInitializationParameter.Value.Value; }
|
|---|
| 195 | set { RandomInitializationParameter.Value.Value = value; }
|
|---|
| 196 | }
|
|---|
| 197 |
|
|---|
| 198 | public int UpdateInterval {
|
|---|
| 199 | get { return UpdateIntervalParameter.Value.Value; }
|
|---|
| 200 | set { UpdateIntervalParameter.Value.Value = value; }
|
|---|
| 201 | }
|
|---|
| 202 | #endregion
|
|---|
| 203 |
|
|---|
| 204 | #region Constructors & Cloning
|
|---|
| 205 | [StorableConstructor]
|
|---|
| 206 | private TSNEAlgorithm(bool deserializing) : base(deserializing) { }
|
|---|
| 207 |
|
|---|
| 208 | private TSNEAlgorithm(TSNEAlgorithm original, Cloner cloner) : base(original, cloner) {
|
|---|
| 209 | if (original.dataRowNames != null)
|
|---|
| 210 | dataRowNames = new Dictionary<string, List<int>>(original.dataRowNames);
|
|---|
| 211 | if (original.dataRows != null)
|
|---|
| 212 | dataRows = original.dataRows.ToDictionary(kvp => kvp.Key, kvp => cloner.Clone(kvp.Value));
|
|---|
| 213 | if (original.state != null)
|
|---|
| 214 | state = cloner.Clone(original.state);
|
|---|
| 215 | iter = original.iter;
|
|---|
| 216 | }
|
|---|
| 217 | public override IDeepCloneable Clone(Cloner cloner) {
|
|---|
| 218 | return new TSNEAlgorithm(this, cloner);
|
|---|
| 219 | }
|
|---|
| 220 | public TSNEAlgorithm() {
|
|---|
| 221 | var distances = new ItemSet<IDistance<double[]>>(ApplicationManager.Manager.GetInstances<IDistance<double[]>>());
|
|---|
| 222 | Parameters.Add(new ConstrainedValueParameter<IDistance<double[]>>(DistanceFunctionParameterName, "The distance function used to differentiate similar from non-similar points", distances, distances.OfType<EuclideanDistance>().FirstOrDefault()));
|
|---|
| 223 | Parameters.Add(new FixedValueParameter<DoubleValue>(PerplexityParameterName, "Perplexity-parameter of tSNE. Comparable to k in a k-nearest neighbour algorithm. Recommended value is floor(number of points /3) or lower", new DoubleValue(25)));
|
|---|
| 224 | Parameters.Add(new FixedValueParameter<PercentValue>(ThetaParameterName, "Value describing how much appoximated " +
|
|---|
| 225 | "gradients my differ from exact gradients. Set to 0 for exact calculation and in [0,1] otherwise. " +
|
|---|
| 226 | "Appropriate values for theta are between 0.1 and 0.7 (default = 0.5). CAUTION: exact calculation of " +
|
|---|
| 227 | "forces requires building a non-sparse N*N matrix where N is the number of data points. This may " +
|
|---|
| 228 | "exceed memory limitations. The function is designed to run on large (N > 5000) data sets. It may give" +
|
|---|
| 229 | " poor performance on very small data sets(it is better to use a standard t - SNE implementation on such data).", new PercentValue(0)));
|
|---|
| 230 | Parameters.Add(new FixedValueParameter<IntValue>(NewDimensionsParameterName, "Dimensionality of projected space (usually 2 for easy visual analysis)", new IntValue(2)));
|
|---|
| 231 | Parameters.Add(new FixedValueParameter<IntValue>(MaxIterationsParameterName, "Maximum number of iterations for gradient descent.", new IntValue(1000)));
|
|---|
| 232 | Parameters.Add(new FixedValueParameter<IntValue>(StopLyingIterationParameterName, "Number of iterations after which p is no longer approximated.", new IntValue(0)));
|
|---|
| 233 | Parameters.Add(new FixedValueParameter<IntValue>(MomentumSwitchIterationParameterName, "Number of iterations after which the momentum in the gradient descent is switched.", new IntValue(0)));
|
|---|
| 234 | Parameters.Add(new FixedValueParameter<DoubleValue>(InitialMomentumParameterName, "The initial momentum in the gradient descent.", new DoubleValue(0.5)));
|
|---|
| 235 | Parameters.Add(new FixedValueParameter<DoubleValue>(FinalMomentumParameterName, "The final momentum.", new DoubleValue(0.8)));
|
|---|
| 236 | Parameters.Add(new FixedValueParameter<DoubleValue>(EtaParameterName, "Gradient descent learning rate.", new DoubleValue(10)));
|
|---|
| 237 | Parameters.Add(new FixedValueParameter<BoolValue>(SetSeedRandomlyParameterName, "If the seed should be random.", new BoolValue(true)));
|
|---|
| 238 | Parameters.Add(new FixedValueParameter<IntValue>(SeedParameterName, "The seed used if it should not be random.", new IntValue(0)));
|
|---|
| 239 | Parameters.Add(new OptionalConstrainedValueParameter<StringValue>(ClassesNameParameterName, "Name of the column specifying the class lables of each data point. If this is not set training/test is used as labels."));
|
|---|
| 240 | Parameters.Add(new FixedValueParameter<BoolValue>(NormalizationParameterName, "Whether the data should be zero centered and have variance of 1 for each variable, so different scalings are ignored.", new BoolValue(true)));
|
|---|
| 241 | Parameters.Add(new FixedValueParameter<IntValue>(UpdateIntervalParameterName, "The interval after which the results will be updated.", new IntValue(50)));
|
|---|
| 242 | Parameters.Add(new FixedValueParameter<BoolValue>(RandomInitializationParameterName, "Wether data points should be randomly initialized or according to the first 2 dimensions", new BoolValue(true)));
|
|---|
| 243 |
|
|---|
| 244 | Parameters[UpdateIntervalParameterName].Hidden = true;
|
|---|
| 245 |
|
|---|
| 246 | MomentumSwitchIterationParameter.Hidden = true;
|
|---|
| 247 | InitialMomentumParameter.Hidden = true;
|
|---|
| 248 | FinalMomentumParameter.Hidden = true;
|
|---|
| 249 | StopLyingIterationParameter.Hidden = true;
|
|---|
| 250 | EtaParameter.Hidden = false;
|
|---|
| 251 | Problem = new RegressionProblem();
|
|---|
| 252 | }
|
|---|
| 253 | #endregion
|
|---|
| 254 |
|
|---|
| 255 | [Storable]
|
|---|
| 256 | private Dictionary<string, List<int>> dataRowNames;
|
|---|
| 257 | [Storable]
|
|---|
| 258 | private Dictionary<string, ScatterPlotDataRow> dataRows;
|
|---|
| 259 | [Storable]
|
|---|
| 260 | private TSNEStatic<double[]>.TSNEState state;
|
|---|
| 261 | [Storable]
|
|---|
| 262 | private int iter;
|
|---|
| 263 |
|
|---|
| 264 | public override void Prepare() {
|
|---|
| 265 | base.Prepare();
|
|---|
| 266 | dataRowNames = null;
|
|---|
| 267 | dataRows = null;
|
|---|
| 268 | state = null;
|
|---|
| 269 | }
|
|---|
| 270 |
|
|---|
| 271 | protected override void Run(CancellationToken cancellationToken) {
|
|---|
| 272 | var problemData = Problem.ProblemData;
|
|---|
| 273 | // set up and initialize everything if necessary
|
|---|
| 274 | if (state == null) {
|
|---|
| 275 | if (SetSeedRandomly) Seed = new System.Random().Next();
|
|---|
| 276 | var random = new MersenneTwister((uint) Seed);
|
|---|
| 277 | var dataset = problemData.Dataset;
|
|---|
| 278 | var allowedInputVariables = problemData.AllowedInputVariables.ToArray();
|
|---|
| 279 | var allindices = Problem.ProblemData.AllIndices.ToArray();
|
|---|
| 280 | var data = allindices.Select(row => allowedInputVariables.Select(col => dataset.GetDoubleValue(col, row)).ToArray()).ToArray();
|
|---|
| 281 | if (Normalization) data = NormalizeInputData(data);
|
|---|
| 282 | state = TSNEStatic<double[]>.CreateState(data, DistanceFunction, random, NewDimensions, Perplexity, Theta, StopLyingIteration, MomentumSwitchIteration, InitialMomentum, FinalMomentum, Eta, RandomInitialization);
|
|---|
| 283 | SetUpResults(allindices);
|
|---|
| 284 | iter = 0;
|
|---|
| 285 | }
|
|---|
| 286 | for (; iter < MaxIterations && !cancellationToken.IsCancellationRequested; iter++) {
|
|---|
| 287 | if (iter % UpdateInterval == 0)
|
|---|
| 288 | Analyze(state);
|
|---|
| 289 | TSNEStatic<double[]>.Iterate(state);
|
|---|
| 290 | }
|
|---|
| 291 | Analyze(state);
|
|---|
| 292 | dataRowNames = null;
|
|---|
| 293 | dataRows = null;
|
|---|
| 294 | state = null;
|
|---|
| 295 | }
|
|---|
| 296 |
|
|---|
| 297 | #region Events
|
|---|
| 298 | protected override void OnProblemChanged() {
|
|---|
| 299 | base.OnProblemChanged();
|
|---|
| 300 | if (Problem == null) return;
|
|---|
| 301 | OnProblemDataChanged(this, null);
|
|---|
| 302 | }
|
|---|
| 303 |
|
|---|
| 304 | protected override void RegisterProblemEvents() {
|
|---|
| 305 | base.RegisterProblemEvents();
|
|---|
| 306 | Problem.ProblemDataChanged += OnProblemDataChanged;
|
|---|
| 307 | }
|
|---|
| 308 | protected override void DeregisterProblemEvents() {
|
|---|
| 309 | base.DeregisterProblemEvents();
|
|---|
| 310 | Problem.ProblemDataChanged -= OnProblemDataChanged;
|
|---|
| 311 | }
|
|---|
| 312 |
|
|---|
| 313 | private void OnProblemDataChanged(object sender, EventArgs args) {
|
|---|
| 314 | if (Problem == null || Problem.ProblemData == null) return;
|
|---|
| 315 | if (!Parameters.ContainsKey(ClassesNameParameterName)) return;
|
|---|
| 316 | ClassesNameParameter.ValidValues.Clear();
|
|---|
| 317 | foreach (var input in Problem.ProblemData.InputVariables) ClassesNameParameter.ValidValues.Add(input);
|
|---|
| 318 | }
|
|---|
| 319 | #endregion
|
|---|
| 320 |
|
|---|
| 321 | #region Helpers
|
|---|
| 322 | private void SetUpResults(IReadOnlyList<int> allIndices) {
|
|---|
| 323 | if (Results == null) return;
|
|---|
| 324 | var results = Results;
|
|---|
| 325 | dataRowNames = new Dictionary<string, List<int>>();
|
|---|
| 326 | dataRows = new Dictionary<string, ScatterPlotDataRow>();
|
|---|
| 327 | var problemData = Problem.ProblemData;
|
|---|
| 328 |
|
|---|
| 329 | //color datapoints acording to classes variable (be it double or string)
|
|---|
| 330 | if (problemData.Dataset.VariableNames.Contains(ClassesName)) {
|
|---|
| 331 | var classificationData = problemData as ClassificationProblemData;
|
|---|
| 332 | if (classificationData != null && classificationData.TargetVariable.Equals(ClassesName)) {
|
|---|
| 333 | var classNames = classificationData.ClassValues.Zip(classificationData.ClassNames, (v, n) => new {v, n}).ToDictionary(x => x.v, x => x.n);
|
|---|
| 334 | var classes = classificationData.Dataset.GetDoubleValues(classificationData.TargetVariable, allIndices).Select(v => classNames[v]).ToArray();
|
|---|
| 335 | for (var i = 0; i < classes.Length; i++) {
|
|---|
| 336 | if (!dataRowNames.ContainsKey(classes[i])) dataRowNames.Add(classes[i], new List<int>());
|
|---|
| 337 | dataRowNames[classes[i]].Add(i);
|
|---|
| 338 | }
|
|---|
| 339 | }
|
|---|
| 340 | else if (((Dataset) problemData.Dataset).VariableHasType<string>(ClassesName)) {
|
|---|
| 341 | var classes = problemData.Dataset.GetStringValues(ClassesName, allIndices).ToArray();
|
|---|
| 342 | for (var i = 0; i < classes.Length; i++) {
|
|---|
| 343 | if (!dataRowNames.ContainsKey(classes[i])) dataRowNames.Add(classes[i], new List<int>());
|
|---|
| 344 | dataRowNames[classes[i]].Add(i);
|
|---|
| 345 | }
|
|---|
| 346 | }
|
|---|
| 347 | else if (((Dataset) problemData.Dataset).VariableHasType<double>(ClassesName)) {
|
|---|
| 348 | var clusterdata = new Dataset(problemData.Dataset.DoubleVariables, problemData.Dataset.DoubleVariables.Select(v => problemData.Dataset.GetDoubleValues(v, allIndices).ToList()));
|
|---|
| 349 | const int contours = 8;
|
|---|
| 350 | Dictionary<int, string> contourMap;
|
|---|
| 351 | IClusteringModel clusterModel;
|
|---|
| 352 | double[][] borders;
|
|---|
| 353 | CreateClusters(clusterdata, ClassesName, contours, out clusterModel, out contourMap, out borders);
|
|---|
| 354 | var contourorder = borders.Select((x, i) => new {x, i}).OrderBy(x => x.x[0]).Select(x => x.i).ToArray();
|
|---|
| 355 | for (var i = 0; i < contours; i++) {
|
|---|
| 356 | var c = contourorder[i];
|
|---|
| 357 | var contourname = contourMap[c];
|
|---|
| 358 | dataRowNames.Add(contourname, new List<int>());
|
|---|
| 359 | dataRows.Add(contourname, new ScatterPlotDataRow(contourname, "", new List<Point2D<double>>()));
|
|---|
| 360 | dataRows[contourname].VisualProperties.Color = GetHeatMapColor(i, contours);
|
|---|
| 361 | dataRows[contourname].VisualProperties.PointSize = i + 3;
|
|---|
| 362 | }
|
|---|
| 363 | var allClusters = clusterModel.GetClusterValues(clusterdata, Enumerable.Range(0, clusterdata.Rows)).ToArray();
|
|---|
| 364 | for (var i = 0; i < clusterdata.Rows; i++) dataRowNames[contourMap[allClusters[i] - 1]].Add(i);
|
|---|
| 365 | }
|
|---|
| 366 | else if (((Dataset) problemData.Dataset).VariableHasType<DateTime>(ClassesName)) {
|
|---|
| 367 | var clusterdata = new Dataset(problemData.Dataset.DateTimeVariables, problemData.Dataset.DateTimeVariables.Select(v => problemData.Dataset.GetDoubleValues(v, allIndices).ToList()));
|
|---|
| 368 | const int contours = 8;
|
|---|
| 369 | Dictionary<int, string> contourMap;
|
|---|
| 370 | IClusteringModel clusterModel;
|
|---|
| 371 | double[][] borders;
|
|---|
| 372 | CreateClusters(clusterdata, ClassesName, contours, out clusterModel, out contourMap, out borders);
|
|---|
| 373 | var contourorder = borders.Select((x, i) => new {x, i}).OrderBy(x => x.x[0]).Select(x => x.i).ToArray();
|
|---|
| 374 | for (var i = 0; i < contours; i++) {
|
|---|
| 375 | var c = contourorder[i];
|
|---|
| 376 | var contourname = contourMap[c];
|
|---|
| 377 | dataRowNames.Add(contourname, new List<int>());
|
|---|
| 378 | dataRows.Add(contourname, new ScatterPlotDataRow(contourname, "", new List<Point2D<double>>()));
|
|---|
| 379 | dataRows[contourname].VisualProperties.Color = GetHeatMapColor(i, contours);
|
|---|
| 380 | dataRows[contourname].VisualProperties.PointSize = i + 3;
|
|---|
| 381 | }
|
|---|
| 382 | var allClusters = clusterModel.GetClusterValues(clusterdata, Enumerable.Range(0, clusterdata.Rows)).ToArray();
|
|---|
| 383 | for (var i = 0; i < clusterdata.Rows; i++) dataRowNames[contourMap[allClusters[i] - 1]].Add(i);
|
|---|
| 384 | }
|
|---|
| 385 | else {
|
|---|
| 386 | dataRowNames.Add("Training", problemData.TrainingIndices.ToList());
|
|---|
| 387 | dataRowNames.Add("Test", problemData.TestIndices.ToList());
|
|---|
| 388 | }
|
|---|
| 389 |
|
|---|
| 390 | if (!results.ContainsKey(IterationResultName)) results.Add(new Result(IterationResultName, new IntValue(0)));
|
|---|
| 391 | else ((IntValue) results[IterationResultName].Value).Value = 0;
|
|---|
| 392 |
|
|---|
| 393 | if (!results.ContainsKey(ErrorResultName)) results.Add(new Result(ErrorResultName, new DoubleValue(0)));
|
|---|
| 394 | else ((DoubleValue) results[ErrorResultName].Value).Value = 0;
|
|---|
| 395 |
|
|---|
| 396 | if (!results.ContainsKey(ErrorPlotResultName)) results.Add(new Result(ErrorPlotResultName, new DataTable(ErrorPlotResultName, "Development of errors during gradient descent")));
|
|---|
| 397 | else results[ErrorPlotResultName].Value = new DataTable(ErrorPlotResultName, "Development of errors during gradient descent");
|
|---|
| 398 |
|
|---|
| 399 | var plot = results[ErrorPlotResultName].Value as DataTable;
|
|---|
| 400 | if (plot == null) throw new ArgumentException("could not create/access error data table in results collection");
|
|---|
| 401 |
|
|---|
| 402 | if (!plot.Rows.ContainsKey("errors")) plot.Rows.Add(new DataRow("errors"));
|
|---|
| 403 | plot.Rows["errors"].Values.Clear();
|
|---|
| 404 | plot.Rows["errors"].VisualProperties.StartIndexZero = true;
|
|---|
| 405 |
|
|---|
| 406 | results.Add(new Result(ScatterPlotResultName, "Plot of the projected data", new ScatterPlot(DataResultName, "")));
|
|---|
| 407 | results.Add(new Result(DataResultName, "Projected Data", new DoubleMatrix()));
|
|---|
| 408 | }
|
|---|
| 409 | }
|
|---|
| 410 |
|
|---|
| 411 | private void Analyze(TSNEStatic<double[]>.TSNEState tsneState) {
|
|---|
| 412 | if (Results == null) return;
|
|---|
| 413 | var results = Results;
|
|---|
| 414 | var plot = results[ErrorPlotResultName].Value as DataTable;
|
|---|
| 415 | if (plot == null) throw new ArgumentException("Could not create/access error data table in results collection.");
|
|---|
| 416 | var errors = plot.Rows["errors"].Values;
|
|---|
| 417 | var c = tsneState.EvaluateError();
|
|---|
| 418 | errors.Add(c);
|
|---|
| 419 | ((IntValue) results[IterationResultName].Value).Value = tsneState.iter;
|
|---|
| 420 | ((DoubleValue) results[ErrorResultName].Value).Value = errors.Last();
|
|---|
| 421 |
|
|---|
| 422 | var ndata = NormalizeProjectedData(tsneState.newData);
|
|---|
| 423 | results[DataResultName].Value = new DoubleMatrix(ndata);
|
|---|
| 424 | var splot = results[ScatterPlotResultName].Value as ScatterPlot;
|
|---|
| 425 | FillScatterPlot(ndata, splot);
|
|---|
| 426 | }
|
|---|
| 427 |
|
|---|
| 428 | private void FillScatterPlot(double[,] lowDimData, ScatterPlot plot) {
|
|---|
| 429 | foreach (var rowName in dataRowNames.Keys) {
|
|---|
| 430 | if (!plot.Rows.ContainsKey(rowName)) {
|
|---|
| 431 | plot.Rows.Add(dataRows.ContainsKey(rowName) ? dataRows[rowName] : new ScatterPlotDataRow(rowName, "", new List<Point2D<double>>()));
|
|---|
| 432 | plot.Rows[rowName].VisualProperties.PointSize = 6;
|
|---|
| 433 | }
|
|---|
| 434 | plot.Rows[rowName].Points.Replace(dataRowNames[rowName].Select(i => new Point2D<double>(lowDimData[i, 0], lowDimData[i, 1])));
|
|---|
| 435 | }
|
|---|
| 436 | }
|
|---|
| 437 |
|
|---|
| 438 | private static double[,] NormalizeProjectedData(double[,] data) {
|
|---|
| 439 | var max = new double[data.GetLength(1)];
|
|---|
| 440 | var min = new double[data.GetLength(1)];
|
|---|
| 441 | var res = new double[data.GetLength(0), data.GetLength(1)];
|
|---|
| 442 | for (var i = 0; i < max.Length; i++) max[i] = min[i] = data[0, i];
|
|---|
| 443 | for (var i = 0; i < data.GetLength(0); i++)
|
|---|
| 444 | for (var j = 0; j < data.GetLength(1); j++) {
|
|---|
| 445 | var v = data[i, j];
|
|---|
| 446 | max[j] = Math.Max(max[j], v);
|
|---|
| 447 | min[j] = Math.Min(min[j], v);
|
|---|
| 448 | }
|
|---|
| 449 | for (var i = 0; i < data.GetLength(0); i++) {
|
|---|
| 450 | for (var j = 0; j < data.GetLength(1); j++) {
|
|---|
| 451 | var d = max[j] - min[j];
|
|---|
| 452 | var s = data[i, j] - (max[j] + min[j]) / 2; //shift data
|
|---|
| 453 | if (d.IsAlmost(0)) res[i, j] = data[i, j]; //no scaling possible
|
|---|
| 454 | else res[i, j] = s / d; //scale data
|
|---|
| 455 | }
|
|---|
| 456 | }
|
|---|
| 457 | return res;
|
|---|
| 458 | }
|
|---|
| 459 |
|
|---|
| 460 | private static double[][] NormalizeInputData(IReadOnlyList<IReadOnlyList<double>> data) {
|
|---|
| 461 | // as in tSNE implementation by van der Maaten
|
|---|
| 462 | var n = data[0].Count;
|
|---|
| 463 | var mean = new double[n];
|
|---|
| 464 | var max = new double[n];
|
|---|
| 465 | var nData = new double[data.Count][];
|
|---|
| 466 | for (var i = 0; i < n; i++) {
|
|---|
| 467 | mean[i] = Enumerable.Range(0, data.Count).Select(x => data[x][i]).Average();
|
|---|
| 468 | max[i] = Enumerable.Range(0, data.Count).Max(x => Math.Abs(data[x][i]));
|
|---|
| 469 | }
|
|---|
| 470 | for (var i = 0; i < data.Count; i++) {
|
|---|
| 471 | nData[i] = new double[n];
|
|---|
| 472 | for (var j = 0; j < n; j++) nData[i][j] = max[j].IsAlmost(0) ? data[i][j] - mean[j] : (data[i][j] - mean[j]) / max[j];
|
|---|
| 473 | }
|
|---|
| 474 | return nData;
|
|---|
| 475 | }
|
|---|
| 476 |
|
|---|
| 477 | private static Color GetHeatMapColor(int contourNr, int noContours) {
|
|---|
| 478 | return ConvertTotalToRgb(0, noContours, contourNr);
|
|---|
| 479 | }
|
|---|
| 480 |
|
|---|
| 481 | private static void CreateClusters(IDataset data, string target, int contours, out IClusteringModel contourCluster, out Dictionary<int, string> contourNames, out double[][] borders) {
|
|---|
| 482 | var cpd = new ClusteringProblemData((Dataset) data, new[] {target});
|
|---|
| 483 | contourCluster = KMeansClustering.CreateKMeansSolution(cpd, contours, 3).Model;
|
|---|
| 484 |
|
|---|
| 485 | borders = Enumerable.Range(0, contours).Select(x => new[] {double.MaxValue, double.MinValue}).ToArray();
|
|---|
| 486 | var clusters = contourCluster.GetClusterValues(cpd.Dataset, cpd.AllIndices).ToArray();
|
|---|
| 487 | var targetvalues = cpd.Dataset.GetDoubleValues(target).ToArray();
|
|---|
| 488 | foreach (var i in cpd.AllIndices) {
|
|---|
| 489 | var cl = clusters[i] - 1;
|
|---|
| 490 | var clv = targetvalues[i];
|
|---|
| 491 | if (borders[cl][0] > clv) borders[cl][0] = clv;
|
|---|
| 492 | if (borders[cl][1] < clv) borders[cl][1] = clv;
|
|---|
| 493 | }
|
|---|
| 494 |
|
|---|
| 495 | contourNames = new Dictionary<int, string>();
|
|---|
| 496 | for (var i = 0; i < contours; i++)
|
|---|
| 497 | contourNames.Add(i, "[" + borders[i][0] + ";" + borders[i][1] + "]");
|
|---|
| 498 | }
|
|---|
| 499 |
|
|---|
| 500 | private static Color ConvertTotalToRgb(double low, double high, double cell) {
|
|---|
| 501 | var range = high - low;
|
|---|
| 502 | var h = cell / range;
|
|---|
| 503 | return HsVtoRgb(h * 0.5, 1.0f, 1.0f);
|
|---|
| 504 | }
|
|---|
| 505 |
|
|---|
| 506 | private static Color HsVtoRgb(double hue, double saturation, double value) {
|
|---|
| 507 | while (hue > 1f) { hue -= 1f; }
|
|---|
| 508 | while (hue < 0f) { hue += 1f; }
|
|---|
| 509 | while (saturation > 1f) { saturation -= 1f; }
|
|---|
| 510 | while (saturation < 0f) { saturation += 1f; }
|
|---|
| 511 | while (value > 1f) { value -= 1f; }
|
|---|
| 512 | while (value < 0f) { value += 1f; }
|
|---|
| 513 | if (hue > 0.999f) { hue = 0.999f; }
|
|---|
| 514 | if (hue < 0.001f) { hue = 0.001f; }
|
|---|
| 515 | if (saturation > 0.999f) { saturation = 0.999f; }
|
|---|
| 516 | if (saturation < 0.001f) { return Color.FromArgb((int) (value * 255f), (int) (value * 255f), (int) (value * 255f)); }
|
|---|
| 517 | if (value > 0.999f) { value = 0.999f; }
|
|---|
| 518 | if (value < 0.001f) { value = 0.001f; }
|
|---|
| 519 |
|
|---|
| 520 | var h6 = hue * 6f;
|
|---|
| 521 | if (h6.IsAlmost(6f)) { h6 = 0f; }
|
|---|
| 522 | var ihue = (int) h6;
|
|---|
| 523 | var p = value * (1f - saturation);
|
|---|
| 524 | var q = value * (1f - saturation * (h6 - ihue));
|
|---|
| 525 | var t = value * (1f - saturation * (1f - (h6 - ihue)));
|
|---|
| 526 | switch (ihue) {
|
|---|
| 527 | case 0:
|
|---|
| 528 | return Color.FromArgb((int) (value * 255), (int) (t * 255), (int) (p * 255));
|
|---|
| 529 | case 1:
|
|---|
| 530 | return Color.FromArgb((int) (q * 255), (int) (value * 255), (int) (p * 255));
|
|---|
| 531 | case 2:
|
|---|
| 532 | return Color.FromArgb((int) (p * 255), (int) (value * 255), (int) (t * 255));
|
|---|
| 533 | case 3:
|
|---|
| 534 | return Color.FromArgb((int) (p * 255), (int) (q * 255), (int) (value * 255));
|
|---|
| 535 | case 4:
|
|---|
| 536 | return Color.FromArgb((int) (t * 255), (int) (p * 255), (int) (value * 255));
|
|---|
| 537 | default:
|
|---|
| 538 | return Color.FromArgb((int) (value * 255), (int) (p * 255), (int) (q * 255));
|
|---|
| 539 | }
|
|---|
| 540 | }
|
|---|
| 541 | #endregion
|
|---|
| 542 | }
|
|---|
| 543 | } |
|---|
