[14386] | 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.Linq;
|
---|
[14872] | 24 | using System.Threading;
|
---|
[14386] | 25 | using HeuristicLab.Common;
|
---|
| 26 | using HeuristicLab.Core;
|
---|
| 27 | using HeuristicLab.Data;
|
---|
| 28 | using HeuristicLab.Optimization;
|
---|
| 29 | using HeuristicLab.Parameters;
|
---|
| 30 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
[15249] | 31 | using HeuristicLab.PluginInfrastructure;
|
---|
[14386] | 32 | using HeuristicLab.Problems.DataAnalysis;
|
---|
| 33 |
|
---|
[15249] | 34 | namespace HeuristicLab.Algorithms.DataAnalysis {
|
---|
[14887] | 35 | [Item("Kernel Ridge Regression", "Kernelized ridge regression e.g. for radial basis function (RBF) regression.")]
|
---|
[14386] | 36 | [Creatable(CreatableAttribute.Categories.DataAnalysisRegression, Priority = 100)]
|
---|
| 37 | [StorableClass]
|
---|
[14887] | 38 | public sealed class KernelRidgeRegression : BasicAlgorithm {
|
---|
| 39 | private const string SolutionResultName = "Kernel ridge regression solution";
|
---|
[14386] | 40 |
|
---|
[14872] | 41 | public override bool SupportsPause {
|
---|
| 42 | get { return false; }
|
---|
| 43 | }
|
---|
| 44 | public override Type ProblemType {
|
---|
| 45 | get { return typeof(IRegressionProblem); }
|
---|
| 46 | }
|
---|
| 47 | public new IRegressionProblem Problem {
|
---|
| 48 | get { return (IRegressionProblem)base.Problem; }
|
---|
| 49 | set { base.Problem = value; }
|
---|
| 50 | }
|
---|
[14386] | 51 |
|
---|
[14872] | 52 | #region parameter names
|
---|
| 53 | private const string KernelParameterName = "Kernel";
|
---|
| 54 | private const string ScaleInputVariablesParameterName = "ScaleInputVariables";
|
---|
[14887] | 55 | private const string LambdaParameterName = "LogLambda";
|
---|
| 56 | private const string BetaParameterName = "Beta";
|
---|
[14386] | 57 | #endregion
|
---|
| 58 |
|
---|
[14872] | 59 | #region parameter properties
|
---|
[15249] | 60 | public IConstrainedValueParameter<IKernel> KernelParameter {
|
---|
| 61 | get { return (IConstrainedValueParameter<IKernel>)Parameters[KernelParameterName]; }
|
---|
[14386] | 62 | }
|
---|
[14872] | 63 |
|
---|
| 64 | public IFixedValueParameter<BoolValue> ScaleInputVariablesParameter {
|
---|
| 65 | get { return (IFixedValueParameter<BoolValue>)Parameters[ScaleInputVariablesParameterName]; }
|
---|
| 66 | }
|
---|
[14887] | 67 |
|
---|
| 68 | public IFixedValueParameter<DoubleValue> LogLambdaParameter {
|
---|
| 69 | get { return (IFixedValueParameter<DoubleValue>)Parameters[LambdaParameterName]; }
|
---|
| 70 | }
|
---|
| 71 |
|
---|
| 72 | public IFixedValueParameter<DoubleValue> BetaParameter {
|
---|
| 73 | get { return (IFixedValueParameter<DoubleValue>)Parameters[BetaParameterName]; }
|
---|
| 74 | }
|
---|
[14386] | 75 | #endregion
|
---|
| 76 |
|
---|
[14872] | 77 | #region properties
|
---|
[14887] | 78 | public IKernel Kernel {
|
---|
[14386] | 79 | get { return KernelParameter.Value; }
|
---|
| 80 | }
|
---|
| 81 |
|
---|
[14872] | 82 | public bool ScaleInputVariables {
|
---|
| 83 | get { return ScaleInputVariablesParameter.Value.Value; }
|
---|
| 84 | set { ScaleInputVariablesParameter.Value.Value = value; }
|
---|
| 85 | }
|
---|
| 86 |
|
---|
[14887] | 87 | public double LogLambda {
|
---|
| 88 | get { return LogLambdaParameter.Value.Value; }
|
---|
| 89 | set { LogLambdaParameter.Value.Value = value; }
|
---|
| 90 | }
|
---|
| 91 |
|
---|
| 92 | public double Beta {
|
---|
| 93 | get { return BetaParameter.Value.Value; }
|
---|
| 94 | set { BetaParameter.Value.Value = value; }
|
---|
| 95 | }
|
---|
[14386] | 96 | #endregion
|
---|
| 97 |
|
---|
| 98 | [StorableConstructor]
|
---|
[14887] | 99 | private KernelRidgeRegression(bool deserializing) : base(deserializing) { }
|
---|
| 100 | private KernelRidgeRegression(KernelRidgeRegression original, Cloner cloner)
|
---|
[14386] | 101 | : base(original, cloner) {
|
---|
| 102 | }
|
---|
[14887] | 103 | public KernelRidgeRegression() {
|
---|
[14386] | 104 | Problem = new RegressionProblem();
|
---|
[15249] | 105 | var values = new ItemSet<IKernel>(ApplicationManager.Manager.GetInstances<IKernel>());
|
---|
| 106 | Parameters.Add(new ConstrainedValueParameter<IKernel>(KernelParameterName, "The kernel", values, values.OfType<GaussianKernel>().FirstOrDefault()));
|
---|
[14872] | 107 | Parameters.Add(new FixedValueParameter<BoolValue>(ScaleInputVariablesParameterName, "Set to true if the input variables should be scaled to the interval [0..1]", new BoolValue(true)));
|
---|
[14887] | 108 | Parameters.Add(new FixedValueParameter<DoubleValue>(LambdaParameterName, "The log10-transformed weight for the regularization term lambda [-inf..+inf]. Small values produce more complex models, large values produce models with larger errors. Set to very small value (e.g. -1.0e15) for almost exact approximation", new DoubleValue(-2)));
|
---|
[15249] | 109 | Parameters.Add(new FixedValueParameter<DoubleValue>(BetaParameterName, "The inverse width of the kernel ]0..+inf]. The distance between points is divided by this value before being plugged into the kernel.", new DoubleValue(2)));
|
---|
[14386] | 110 | }
|
---|
| 111 |
|
---|
| 112 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
[14887] | 113 | return new KernelRidgeRegression(this, cloner);
|
---|
[14386] | 114 | }
|
---|
| 115 |
|
---|
[14872] | 116 | protected override void Run(CancellationToken cancellationToken) {
|
---|
[14888] | 117 | double rmsError, looCvRMSE;
|
---|
[14887] | 118 | var kernel = Kernel;
|
---|
| 119 | kernel.Beta = Beta;
|
---|
[14888] | 120 | var solution = CreateRadialBasisRegressionSolution(Problem.ProblemData, kernel, Math.Pow(10, LogLambda), ScaleInputVariables, out rmsError, out looCvRMSE);
|
---|
[14887] | 121 | Results.Add(new Result(SolutionResultName, "The kernel ridge regression solution.", solution));
|
---|
[14872] | 122 | Results.Add(new Result("RMSE (test)", "The root mean squared error of the solution on the test set.", new DoubleValue(rmsError)));
|
---|
[14888] | 123 | Results.Add(new Result("RMSE (LOO-CV)", "The leave-one-out-cross-validation root mean squared error", new DoubleValue(looCvRMSE)));
|
---|
[14386] | 124 | }
|
---|
| 125 |
|
---|
[14888] | 126 | public static IRegressionSolution CreateRadialBasisRegressionSolution(IRegressionProblemData problemData, ICovarianceFunction kernel, double lambda, bool scaleInputs, out double rmsError, out double looCvRMSE) {
|
---|
[15249] | 127 | var model = KernelRidgeRegressionModel.Create(problemData.Dataset, problemData.TargetVariable, problemData.AllowedInputVariables, problemData.TrainingIndices, scaleInputs, kernel, lambda);
|
---|
[14887] | 128 | rmsError = double.NaN;
|
---|
| 129 | if (problemData.TestIndices.Any()) {
|
---|
| 130 | rmsError = Math.Sqrt(model.GetEstimatedValues(problemData.Dataset, problemData.TestIndices)
|
---|
| 131 | .Zip(problemData.TargetVariableTestValues, (a, b) => (a - b) * (a - b))
|
---|
| 132 | .Average());
|
---|
| 133 | }
|
---|
[14872] | 134 | var solution = model.CreateRegressionSolution((IRegressionProblemData)problemData.Clone());
|
---|
[14887] | 135 | solution.Model.Name = "Kernel ridge regression model";
|
---|
| 136 | solution.Name = SolutionResultName;
|
---|
[14888] | 137 | looCvRMSE = model.LooCvRMSE;
|
---|
[14386] | 138 | return solution;
|
---|
| 139 | }
|
---|
| 140 | }
|
---|
| 141 | }
|
---|