1 | #region License Information
|
---|
2 | /* HeuristicLab
|
---|
3 | * Copyright (C) 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.Linq;
|
---|
24 | using System.Threading;
|
---|
25 | using HEAL.Attic;
|
---|
26 | using HeuristicLab.Common;
|
---|
27 | using HeuristicLab.Core;
|
---|
28 | using HeuristicLab.Data;
|
---|
29 | using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
|
---|
30 | using HeuristicLab.Optimization;
|
---|
31 | using HeuristicLab.Parameters;
|
---|
32 | using HeuristicLab.Problems.DataAnalysis;
|
---|
33 | using HeuristicLab.Problems.Instances;
|
---|
34 |
|
---|
35 |
|
---|
36 | namespace HeuristicLab.Algorithms.DataAnalysis {
|
---|
37 | [Item("Gaussian Process Covariance Optimization Problem", "")]
|
---|
38 | [Creatable(CreatableAttribute.Categories.GeneticProgrammingProblems, Priority = 300)]
|
---|
39 | [StorableType("A3EA7CE7-78FA-48FF-9DD5-FBE5AB770A99")]
|
---|
40 | public sealed class GaussianProcessCovarianceOptimizationProblem : SymbolicExpressionTreeProblem, IStatefulItem, IRegressionProblem, IProblemInstanceConsumer<IRegressionProblemData>, IProblemInstanceExporter<IRegressionProblemData> {
|
---|
41 | #region static variables and ctor
|
---|
42 | private static readonly CovarianceMaternIso maternIso1;
|
---|
43 | private static readonly CovarianceMaternIso maternIso3;
|
---|
44 | private static readonly CovarianceMaternIso maternIso5;
|
---|
45 | private static readonly CovariancePiecewisePolynomial piecewisePoly0;
|
---|
46 | private static readonly CovariancePiecewisePolynomial piecewisePoly1;
|
---|
47 | private static readonly CovariancePiecewisePolynomial piecewisePoly2;
|
---|
48 | private static readonly CovariancePiecewisePolynomial piecewisePoly3;
|
---|
49 | private static readonly CovariancePolynomial poly2;
|
---|
50 | private static readonly CovariancePolynomial poly3;
|
---|
51 | private static readonly CovarianceSpectralMixture spectralMixture1;
|
---|
52 | private static readonly CovarianceSpectralMixture spectralMixture3;
|
---|
53 | private static readonly CovarianceSpectralMixture spectralMixture5;
|
---|
54 | private static readonly CovarianceLinear linear;
|
---|
55 | private static readonly CovarianceLinearArd linearArd;
|
---|
56 | private static readonly CovarianceNeuralNetwork neuralNetwork;
|
---|
57 | private static readonly CovariancePeriodic periodic;
|
---|
58 | private static readonly CovarianceRationalQuadraticIso ratQuadraticIso;
|
---|
59 | private static readonly CovarianceRationalQuadraticArd ratQuadraticArd;
|
---|
60 | private static readonly CovarianceSquaredExponentialArd sqrExpArd;
|
---|
61 | private static readonly CovarianceSquaredExponentialIso sqrExpIso;
|
---|
62 |
|
---|
63 | static GaussianProcessCovarianceOptimizationProblem() {
|
---|
64 | // cumbersome initialization because of ConstrainedValueParameters
|
---|
65 | maternIso1 = new CovarianceMaternIso(); SetConstrainedValueParameter(maternIso1.DParameter, 1);
|
---|
66 | maternIso3 = new CovarianceMaternIso(); SetConstrainedValueParameter(maternIso3.DParameter, 3);
|
---|
67 | maternIso5 = new CovarianceMaternIso(); SetConstrainedValueParameter(maternIso5.DParameter, 5);
|
---|
68 |
|
---|
69 | piecewisePoly0 = new CovariancePiecewisePolynomial(); SetConstrainedValueParameter(piecewisePoly0.VParameter, 0);
|
---|
70 | piecewisePoly1 = new CovariancePiecewisePolynomial(); SetConstrainedValueParameter(piecewisePoly1.VParameter, 1);
|
---|
71 | piecewisePoly2 = new CovariancePiecewisePolynomial(); SetConstrainedValueParameter(piecewisePoly2.VParameter, 2);
|
---|
72 | piecewisePoly3 = new CovariancePiecewisePolynomial(); SetConstrainedValueParameter(piecewisePoly3.VParameter, 3);
|
---|
73 |
|
---|
74 | poly2 = new CovariancePolynomial(); poly2.DegreeParameter.Value.Value = 2;
|
---|
75 | poly3 = new CovariancePolynomial(); poly3.DegreeParameter.Value.Value = 3;
|
---|
76 |
|
---|
77 | spectralMixture1 = new CovarianceSpectralMixture(); spectralMixture1.QParameter.Value.Value = 1;
|
---|
78 | spectralMixture3 = new CovarianceSpectralMixture(); spectralMixture3.QParameter.Value.Value = 3;
|
---|
79 | spectralMixture5 = new CovarianceSpectralMixture(); spectralMixture5.QParameter.Value.Value = 5;
|
---|
80 |
|
---|
81 | linear = new CovarianceLinear();
|
---|
82 | linearArd = new CovarianceLinearArd();
|
---|
83 | neuralNetwork = new CovarianceNeuralNetwork();
|
---|
84 | periodic = new CovariancePeriodic();
|
---|
85 | ratQuadraticArd = new CovarianceRationalQuadraticArd();
|
---|
86 | ratQuadraticIso = new CovarianceRationalQuadraticIso();
|
---|
87 | sqrExpArd = new CovarianceSquaredExponentialArd();
|
---|
88 | sqrExpIso = new CovarianceSquaredExponentialIso();
|
---|
89 | }
|
---|
90 |
|
---|
91 | private static void SetConstrainedValueParameter(IConstrainedValueParameter<IntValue> param, int val) {
|
---|
92 | param.Value = param.ValidValues.Single(v => v.Value == val);
|
---|
93 | }
|
---|
94 |
|
---|
95 | #endregion
|
---|
96 |
|
---|
97 | #region parameter names
|
---|
98 |
|
---|
99 | private const string ProblemDataParameterName = "ProblemData";
|
---|
100 | private const string ConstantOptIterationsParameterName = "Constant optimization steps";
|
---|
101 | private const string RestartsParameterName = "Restarts";
|
---|
102 |
|
---|
103 | #endregion
|
---|
104 |
|
---|
105 | #region Parameter Properties
|
---|
106 | public IValueParameter<IRegressionProblemData> ProblemDataParameter {
|
---|
107 | get { return (IValueParameter<IRegressionProblemData>)Parameters[ProblemDataParameterName]; }
|
---|
108 | }
|
---|
109 | public IFixedValueParameter<IntValue> ConstantOptIterationsParameter {
|
---|
110 | get { return (IFixedValueParameter<IntValue>)Parameters[ConstantOptIterationsParameterName]; }
|
---|
111 | }
|
---|
112 | public IFixedValueParameter<IntValue> RestartsParameter {
|
---|
113 | get { return (IFixedValueParameter<IntValue>)Parameters[RestartsParameterName]; }
|
---|
114 | }
|
---|
115 | #endregion
|
---|
116 |
|
---|
117 | #region Properties
|
---|
118 |
|
---|
119 | public IRegressionProblemData ProblemData {
|
---|
120 | get { return ProblemDataParameter.Value; }
|
---|
121 | set { ProblemDataParameter.Value = value; }
|
---|
122 | }
|
---|
123 | IDataAnalysisProblemData IDataAnalysisProblem.ProblemData { get { return ProblemData; } }
|
---|
124 |
|
---|
125 | public int ConstantOptIterations {
|
---|
126 | get { return ConstantOptIterationsParameter.Value.Value; }
|
---|
127 | set { ConstantOptIterationsParameter.Value.Value = value; }
|
---|
128 | }
|
---|
129 |
|
---|
130 | public int Restarts {
|
---|
131 | get { return RestartsParameter.Value.Value; }
|
---|
132 | set { RestartsParameter.Value.Value = value; }
|
---|
133 | }
|
---|
134 | #endregion
|
---|
135 |
|
---|
136 | // problem stores a few variables for information exchange from Evaluate() to Analyze()
|
---|
137 | private readonly object problemStateLocker = new object();
|
---|
138 | [Storable]
|
---|
139 | private double bestQ;
|
---|
140 | [Storable]
|
---|
141 | private double[] bestHyperParameters;
|
---|
142 | [Storable]
|
---|
143 | private IMeanFunction meanFunc;
|
---|
144 | [Storable]
|
---|
145 | private ICovarianceFunction covFunc;
|
---|
146 |
|
---|
147 | public GaussianProcessCovarianceOptimizationProblem() : base(new SymbolicExpressionTreeEncoding()) {
|
---|
148 | Maximization = true; // return log likelihood (instead of negative log likelihood as in GPR)
|
---|
149 | Parameters.Add(new ValueParameter<IRegressionProblemData>(ProblemDataParameterName, "The data for the regression problem", new RegressionProblemData()));
|
---|
150 | Parameters.Add(new FixedValueParameter<IntValue>(ConstantOptIterationsParameterName, "Number of optimization steps for hyperparameter values", new IntValue(50)));
|
---|
151 | Parameters.Add(new FixedValueParameter<IntValue>(RestartsParameterName, "The number of random restarts for constant optimization.", new IntValue(10)));
|
---|
152 | Parameters["Restarts"].Hidden = true;
|
---|
153 |
|
---|
154 |
|
---|
155 | var g = new SimpleSymbolicExpressionGrammar();
|
---|
156 | g.AddSymbols(new string[] { "Sum", "Product" }, 2, 2);
|
---|
157 | g.AddTerminalSymbols(new string[]
|
---|
158 | {
|
---|
159 | "Linear",
|
---|
160 | "LinearArd",
|
---|
161 | "MaternIso1",
|
---|
162 | "MaternIso3",
|
---|
163 | "MaternIso5",
|
---|
164 | "NeuralNetwork",
|
---|
165 | "Periodic",
|
---|
166 | "PiecewisePolynomial0",
|
---|
167 | "PiecewisePolynomial1",
|
---|
168 | "PiecewisePolynomial2",
|
---|
169 | "PiecewisePolynomial3",
|
---|
170 | "Polynomial2",
|
---|
171 | "Polynomial3",
|
---|
172 | "RationalQuadraticArd",
|
---|
173 | "RationalQuadraticIso",
|
---|
174 | "SpectralMixture1",
|
---|
175 | "SpectralMixture3",
|
---|
176 | "SpectralMixture5",
|
---|
177 | "SquaredExponentialArd",
|
---|
178 | "SquaredExponentialIso"
|
---|
179 | });
|
---|
180 |
|
---|
181 | Encoding.TreeLength = 10;
|
---|
182 | Encoding.TreeDepth = 5;
|
---|
183 | Encoding.Grammar = g;
|
---|
184 | Encoding.GrammarParameter.ReadOnly = GrammarRefParameter.ReadOnly = true;
|
---|
185 | }
|
---|
186 |
|
---|
187 | public void InitializeState() { ClearState(); }
|
---|
188 | public void ClearState() {
|
---|
189 | meanFunc = null;
|
---|
190 | covFunc = null;
|
---|
191 | bestQ = double.NegativeInfinity;
|
---|
192 | bestHyperParameters = null;
|
---|
193 | }
|
---|
194 |
|
---|
195 | private readonly object syncRoot = new object();
|
---|
196 | // Does not produce the same result for the same seed when using parallel engine (see below)!
|
---|
197 | public override ISingleObjectiveEvaluationResult Evaluate(ISymbolicExpressionTree tree, IRandom random, CancellationToken cancellationToken) {
|
---|
198 | var meanFunction = new MeanConst();
|
---|
199 | var problemData = ProblemData;
|
---|
200 | var ds = problemData.Dataset;
|
---|
201 | var targetVariable = problemData.TargetVariable;
|
---|
202 | var allowedInputVariables = problemData.AllowedInputVariables.ToArray();
|
---|
203 | var nVars = allowedInputVariables.Length;
|
---|
204 | var trainingRows = problemData.TrainingIndices.ToArray();
|
---|
205 |
|
---|
206 | // use the same covariance function for each restart
|
---|
207 | var covarianceFunction = TreeToCovarianceFunction(tree);
|
---|
208 |
|
---|
209 | // allocate hyperparameters
|
---|
210 | var hyperParameters = new double[meanFunction.GetNumberOfParameters(nVars) + covarianceFunction.GetNumberOfParameters(nVars) + 1]; // mean + cov + noise
|
---|
211 | double[] bestHyperParameters = new double[hyperParameters.Length];
|
---|
212 | var bestObjValue = new double[1] { double.MinValue };
|
---|
213 |
|
---|
214 | // data that is necessary for the objective function
|
---|
215 | var data = Tuple.Create(ds, targetVariable, allowedInputVariables, trainingRows, (IMeanFunction)meanFunction, covarianceFunction, bestObjValue);
|
---|
216 |
|
---|
217 | for (int t = 0; t < Restarts; t++) {
|
---|
218 | var prevBest = bestObjValue[0];
|
---|
219 | var prevBestHyperParameters = new double[hyperParameters.Length];
|
---|
220 | Array.Copy(bestHyperParameters, prevBestHyperParameters, bestHyperParameters.Length);
|
---|
221 |
|
---|
222 | // initialize hyperparameters
|
---|
223 | hyperParameters[0] = ds.GetDoubleValues(targetVariable).Average(); // mean const
|
---|
224 |
|
---|
225 | // Evaluate might be called concurrently therefore access to random has to be synchronized.
|
---|
226 | // However, results of multiple runs with the same seed will be different when using the parallel engine.
|
---|
227 | lock (syncRoot) {
|
---|
228 | for (int i = 0; i < covarianceFunction.GetNumberOfParameters(nVars); i++) {
|
---|
229 | hyperParameters[1 + i] = random.NextDouble() * 2.0 - 1.0;
|
---|
230 | }
|
---|
231 | }
|
---|
232 | hyperParameters[hyperParameters.Length - 1] = 1.0; // s² = exp(2), TODO: other inits better?
|
---|
233 |
|
---|
234 | // use alglib.bfgs for hyper-parameter optimization ...
|
---|
235 | double epsg = 0;
|
---|
236 | double epsf = 0.00001;
|
---|
237 | double epsx = 0;
|
---|
238 | double stpmax = 1;
|
---|
239 | int maxits = ConstantOptIterations;
|
---|
240 | alglib.mincgstate state;
|
---|
241 | alglib.mincgreport rep;
|
---|
242 |
|
---|
243 | alglib.mincgcreate(hyperParameters, out state);
|
---|
244 | alglib.mincgsetcond(state, epsg, epsf, epsx, maxits);
|
---|
245 | alglib.mincgsetstpmax(state, stpmax);
|
---|
246 | alglib.mincgoptimize(state, ObjectiveFunction, null, data);
|
---|
247 |
|
---|
248 | alglib.mincgresults(state, out bestHyperParameters, out rep);
|
---|
249 |
|
---|
250 | if (rep.terminationtype < 0) {
|
---|
251 | // error -> restore previous best quality
|
---|
252 | bestObjValue[0] = prevBest;
|
---|
253 | Array.Copy(prevBestHyperParameters, bestHyperParameters, prevBestHyperParameters.Length);
|
---|
254 | }
|
---|
255 | }
|
---|
256 |
|
---|
257 | UpdateBestSoFar(bestObjValue[0], bestHyperParameters, meanFunction, covarianceFunction);
|
---|
258 |
|
---|
259 | var quality = bestObjValue[0];
|
---|
260 | return new SingleObjectiveEvaluationResult(quality);
|
---|
261 | }
|
---|
262 |
|
---|
263 | // updates the overall best quality and overall best model for Analyze()
|
---|
264 | private void UpdateBestSoFar(double bestQ, double[] bestHyperParameters, IMeanFunction meanFunc, ICovarianceFunction covFunc) {
|
---|
265 | lock (problemStateLocker) {
|
---|
266 | if (bestQ > this.bestQ) {
|
---|
267 | this.bestQ = bestQ;
|
---|
268 | this.bestHyperParameters = new double[bestHyperParameters.Length];
|
---|
269 | Array.Copy(bestHyperParameters, this.bestHyperParameters, this.bestHyperParameters.Length);
|
---|
270 | this.meanFunc = meanFunc;
|
---|
271 | this.covFunc = covFunc;
|
---|
272 | }
|
---|
273 | }
|
---|
274 | }
|
---|
275 |
|
---|
276 |
|
---|
277 | public override void Analyze(ISingleObjectiveSolutionContext<ISymbolicExpressionTree>[] solutionContexts, IRandom random) {
|
---|
278 | //TODO: reimplement code below with results parameter
|
---|
279 |
|
---|
280 | //if (!results.ContainsKey("Best Solution Quality")) {
|
---|
281 | // results.Add(new Result("Best Solution Quality", typeof(DoubleValue)));
|
---|
282 | //}
|
---|
283 | //if (!results.ContainsKey("Best Tree")) {
|
---|
284 | // results.Add(new Result("Best Tree", typeof(ISymbolicExpressionTree)));
|
---|
285 | //}
|
---|
286 | //if (!results.ContainsKey("Best Solution")) {
|
---|
287 | // results.Add(new Result("Best Solution", typeof(GaussianProcessRegressionSolution)));
|
---|
288 | //}
|
---|
289 |
|
---|
290 | //var bestQuality = qualities.Max();
|
---|
291 |
|
---|
292 | //if (results["Best Solution Quality"].Value == null || bestQuality > ((DoubleValue)results["Best Solution Quality"].Value).Value) {
|
---|
293 | // var bestIdx = Array.IndexOf(qualities, bestQuality);
|
---|
294 | // var bestClone = (ISymbolicExpressionTree)trees[bestIdx].Clone();
|
---|
295 | // results["Best Tree"].Value = bestClone;
|
---|
296 | // results["Best Solution Quality"].Value = new DoubleValue(bestQuality);
|
---|
297 | // results["Best Solution"].Value = CreateSolution();
|
---|
298 | //}
|
---|
299 | }
|
---|
300 |
|
---|
301 | private IItem CreateSolution() {
|
---|
302 | var problemData = ProblemData;
|
---|
303 | var ds = problemData.Dataset;
|
---|
304 | var targetVariable = problemData.TargetVariable;
|
---|
305 | var allowedInputVariables = problemData.AllowedInputVariables.ToArray();
|
---|
306 | var trainingRows = problemData.TrainingIndices.ToArray();
|
---|
307 |
|
---|
308 | lock (problemStateLocker) {
|
---|
309 | var model = new GaussianProcessModel(ds, targetVariable, allowedInputVariables, trainingRows, bestHyperParameters, (IMeanFunction)meanFunc.Clone(), (ICovarianceFunction)covFunc.Clone());
|
---|
310 | model.FixParameters();
|
---|
311 | return model.CreateRegressionSolution((IRegressionProblemData)ProblemData.Clone());
|
---|
312 | }
|
---|
313 | }
|
---|
314 |
|
---|
315 | private void ObjectiveFunction(double[] x, ref double func, double[] grad, object obj) {
|
---|
316 | // we want to optimize the model likelihood by changing the hyperparameters and also return the gradient for each hyperparameter
|
---|
317 | var data = (Tuple<IDataset, string, string[], int[], IMeanFunction, ICovarianceFunction, double[]>)obj;
|
---|
318 | var ds = data.Item1;
|
---|
319 | var targetVariable = data.Item2;
|
---|
320 | var allowedInputVariables = data.Item3;
|
---|
321 | var trainingRows = data.Item4;
|
---|
322 | var meanFunction = data.Item5;
|
---|
323 | var covarianceFunction = data.Item6;
|
---|
324 | var bestObjValue = data.Item7;
|
---|
325 | var hyperParameters = x; // the decision variable vector
|
---|
326 |
|
---|
327 | try {
|
---|
328 | var model = new GaussianProcessModel(ds, targetVariable, allowedInputVariables, trainingRows, hyperParameters, meanFunction, covarianceFunction);
|
---|
329 |
|
---|
330 | func = model.NegativeLogLikelihood; // mincgoptimize, so we return negative likelihood
|
---|
331 | bestObjValue[0] = Math.Max(bestObjValue[0], -func); // problem itself is a maximization problem
|
---|
332 | var gradients = model.HyperparameterGradients;
|
---|
333 | Array.Copy(gradients, grad, gradients.Length);
|
---|
334 | } catch (ArgumentException) {
|
---|
335 | // building the GaussianProcessModel might fail, in this case we return the worst possible objective value
|
---|
336 | func = 1.0E+300;
|
---|
337 | Array.Clear(grad, 0, grad.Length);
|
---|
338 | }
|
---|
339 | }
|
---|
340 |
|
---|
341 | private ICovarianceFunction TreeToCovarianceFunction(ISymbolicExpressionTree tree) {
|
---|
342 | return TreeToCovarianceFunction(tree.Root.GetSubtree(0).GetSubtree(0)); // skip programroot and startsymbol
|
---|
343 | }
|
---|
344 |
|
---|
345 | private ICovarianceFunction TreeToCovarianceFunction(ISymbolicExpressionTreeNode node) {
|
---|
346 | switch (node.Symbol.Name) {
|
---|
347 | case "Sum": {
|
---|
348 | var sum = new CovarianceSum();
|
---|
349 | sum.Terms.Add(TreeToCovarianceFunction(node.GetSubtree(0)));
|
---|
350 | sum.Terms.Add(TreeToCovarianceFunction(node.GetSubtree(1)));
|
---|
351 | return sum;
|
---|
352 | }
|
---|
353 | case "Product": {
|
---|
354 | var prod = new CovarianceProduct();
|
---|
355 | prod.Factors.Add(TreeToCovarianceFunction(node.GetSubtree(0)));
|
---|
356 | prod.Factors.Add(TreeToCovarianceFunction(node.GetSubtree(1)));
|
---|
357 | return prod;
|
---|
358 | }
|
---|
359 | // covFunction is cloned by the model so we can reuse instances of terminal covariance functions
|
---|
360 | case "Linear": return linear;
|
---|
361 | case "LinearArd": return linearArd;
|
---|
362 | case "MaternIso1": return maternIso1;
|
---|
363 | case "MaternIso3": return maternIso3;
|
---|
364 | case "MaternIso5": return maternIso5;
|
---|
365 | case "NeuralNetwork": return neuralNetwork;
|
---|
366 | case "Periodic": return periodic;
|
---|
367 | case "PiecewisePolynomial0": return piecewisePoly0;
|
---|
368 | case "PiecewisePolynomial1": return piecewisePoly1;
|
---|
369 | case "PiecewisePolynomial2": return piecewisePoly2;
|
---|
370 | case "PiecewisePolynomial3": return piecewisePoly3;
|
---|
371 | case "Polynomial2": return poly2;
|
---|
372 | case "Polynomial3": return poly3;
|
---|
373 | case "RationalQuadraticArd": return ratQuadraticArd;
|
---|
374 | case "RationalQuadraticIso": return ratQuadraticIso;
|
---|
375 | case "SpectralMixture1": return spectralMixture1;
|
---|
376 | case "SpectralMixture3": return spectralMixture3;
|
---|
377 | case "SpectralMixture5": return spectralMixture5;
|
---|
378 | case "SquaredExponentialArd": return sqrExpArd;
|
---|
379 | case "SquaredExponentialIso": return sqrExpIso;
|
---|
380 | default: throw new InvalidProgramException(string.Format("Found invalid symbol {0}", node.Symbol.Name));
|
---|
381 | }
|
---|
382 | }
|
---|
383 |
|
---|
384 |
|
---|
385 | // persistence
|
---|
386 | [StorableConstructor]
|
---|
387 | private GaussianProcessCovarianceOptimizationProblem(StorableConstructorFlag _) : base(_) { }
|
---|
388 | [StorableHook(HookType.AfterDeserialization)]
|
---|
389 | private void AfterDeserialization() {
|
---|
390 | }
|
---|
391 |
|
---|
392 | // cloning
|
---|
393 | private GaussianProcessCovarianceOptimizationProblem(GaussianProcessCovarianceOptimizationProblem original, Cloner cloner)
|
---|
394 | : base(original, cloner) {
|
---|
395 | bestQ = original.bestQ;
|
---|
396 | meanFunc = cloner.Clone(original.meanFunc);
|
---|
397 | covFunc = cloner.Clone(original.covFunc);
|
---|
398 | if (bestHyperParameters != null) {
|
---|
399 | bestHyperParameters = new double[original.bestHyperParameters.Length];
|
---|
400 | Array.Copy(original.bestHyperParameters, bestHyperParameters, bestHyperParameters.Length);
|
---|
401 | }
|
---|
402 | }
|
---|
403 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
404 | return new GaussianProcessCovarianceOptimizationProblem(this, cloner);
|
---|
405 | }
|
---|
406 |
|
---|
407 | public void Load(IRegressionProblemData data) {
|
---|
408 | this.ProblemData = data;
|
---|
409 | OnProblemDataChanged();
|
---|
410 | }
|
---|
411 |
|
---|
412 | public IRegressionProblemData Export() {
|
---|
413 | return ProblemData;
|
---|
414 | }
|
---|
415 |
|
---|
416 | #region events
|
---|
417 | public event EventHandler ProblemDataChanged;
|
---|
418 |
|
---|
419 |
|
---|
420 | private void OnProblemDataChanged() {
|
---|
421 | var handler = ProblemDataChanged;
|
---|
422 | if (handler != null)
|
---|
423 | handler(this, EventArgs.Empty);
|
---|
424 | }
|
---|
425 | #endregion
|
---|
426 |
|
---|
427 | }
|
---|
428 | }
|
---|