[15064] | 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 HeuristicLab.Common;
|
---|
| 24 | using HeuristicLab.Core;
|
---|
| 25 | using HeuristicLab.Data;
|
---|
| 26 | using HeuristicLab.Encodings.RealVectorEncoding;
|
---|
| 27 | using HeuristicLab.Parameters;
|
---|
| 28 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
| 29 | using HeuristicLab.Problems.DataAnalysis;
|
---|
| 30 |
|
---|
| 31 | // ReSharper disable once CheckNamespace
|
---|
| 32 | namespace HeuristicLab.Algorithms.EGO {
|
---|
| 33 |
|
---|
| 34 | [StorableClass]
|
---|
| 35 | public abstract class ExpectedImprovementBase : InfillCriterionBase {
|
---|
| 36 |
|
---|
| 37 | #region ParameterNames
|
---|
| 38 | private const string ExploitationWeightParameterName = "ExploitationWeight";
|
---|
| 39 | #endregion
|
---|
| 40 |
|
---|
| 41 | #region ParameterProperties
|
---|
| 42 | public IFixedValueParameter<DoubleValue> ExploitationWeightParameter => Parameters[ExploitationWeightParameterName] as IFixedValueParameter<DoubleValue>;
|
---|
| 43 | #endregion
|
---|
| 44 |
|
---|
| 45 | #region Properties
|
---|
| 46 | protected double ExploitationWeight => ExploitationWeightParameter.Value.Value;
|
---|
| 47 | [Storable]
|
---|
| 48 | protected double BestFitness;
|
---|
| 49 | #endregion
|
---|
| 50 |
|
---|
| 51 | #region Constructors, Serialization and Cloning
|
---|
| 52 | [StorableConstructor]
|
---|
| 53 | protected ExpectedImprovementBase(bool deserializing) : base(deserializing) { }
|
---|
| 54 | [StorableHook(HookType.AfterDeserialization)]
|
---|
| 55 | private void AfterDeserialization() {
|
---|
| 56 | RegisterEventhandlers();
|
---|
| 57 | }
|
---|
| 58 | protected ExpectedImprovementBase(ExpectedImprovementBase original, Cloner cloner) : base(original, cloner) {
|
---|
| 59 | BestFitness = original.BestFitness;
|
---|
| 60 | RegisterEventhandlers();
|
---|
| 61 | }
|
---|
| 62 | protected ExpectedImprovementBase() {
|
---|
| 63 | Parameters.Add(new FixedValueParameter<DoubleValue>(ExploitationWeightParameterName, "A value between 0 and 1 indicating the focus on exploration (0) or exploitation (1). 0.5 equates to the original EI by Jones et al.", new DoubleValue(0.5)));
|
---|
| 64 | RegisterEventhandlers();
|
---|
| 65 | }
|
---|
| 66 | #endregion
|
---|
| 67 |
|
---|
| 68 | public override void Initialize() {
|
---|
| 69 | var solution = RegressionSolution as IConfidenceRegressionSolution;
|
---|
| 70 | if (solution == null) throw new ArgumentException("can not calculate EI without a regression solution providing confidence values");
|
---|
| 71 | BestFitness = FindBestFitness(solution);
|
---|
| 72 | }
|
---|
| 73 |
|
---|
| 74 | protected abstract double FindBestFitness(IConfidenceRegressionSolution solution);
|
---|
| 75 |
|
---|
| 76 | public override double Evaluate(RealVector vector) {
|
---|
| 77 | var model = RegressionSolution.Model as IConfidenceRegressionModel;
|
---|
| 78 | var yhat = model.GetEstimation(vector);
|
---|
| 79 | var s = Math.Sqrt(model.GetVariance(vector));
|
---|
| 80 | return Evaluate(vector, yhat, s);
|
---|
| 81 | }
|
---|
| 82 |
|
---|
| 83 | protected abstract double Evaluate(RealVector vector, double estimatedFitness, double estimatedStandardDeviation);
|
---|
| 84 |
|
---|
| 85 | #region Eventhandling
|
---|
| 86 | private void RegisterEventhandlers() {
|
---|
| 87 | DeregisterEventhandlers();
|
---|
| 88 | ExploitationWeightParameter.Value.ValueChanged += OnExploitationWeightChanged;
|
---|
| 89 | }
|
---|
| 90 | private void DeregisterEventhandlers() {
|
---|
| 91 | ExploitationWeightParameter.Value.ValueChanged -= OnExploitationWeightChanged;
|
---|
| 92 | }
|
---|
| 93 | private void OnExploitationWeightChanged(object sender, EventArgs e) {
|
---|
| 94 | ExploitationWeightParameter.Value.ValueChanged -= OnExploitationWeightChanged;
|
---|
| 95 | ExploitationWeightParameter.Value.Value = Math.Max(0, Math.Min(ExploitationWeight, 1));
|
---|
| 96 | ExploitationWeightParameter.Value.ValueChanged += OnExploitationWeightChanged;
|
---|
| 97 | }
|
---|
| 98 | #endregion
|
---|
| 99 |
|
---|
| 100 | #region Helpers
|
---|
| 101 | public static double GetEstimatedImprovement(double bestFitness, double estimatedFitness, double modelUncertainty, double weight, bool maximization) {
|
---|
| 102 | if (Math.Abs(modelUncertainty) < double.Epsilon) return 0;
|
---|
| 103 | var diff = maximization ? (estimatedFitness - bestFitness) : (bestFitness - estimatedFitness);
|
---|
| 104 | var val = diff / modelUncertainty;
|
---|
| 105 | var res = weight * diff * StandardNormalDistribution(val) + (1 - weight) * modelUncertainty * StandardNormalDensity(val);
|
---|
| 106 | return double.IsInfinity(res) || double.IsNaN(res) ? 0 : res;
|
---|
| 107 | }
|
---|
| 108 | private static double StandardNormalDensity(double x) {
|
---|
| 109 | if (Math.Abs(x) > 10) return 0;
|
---|
| 110 | return Math.Exp(-0.5 * x * x) / Math.Sqrt(2 * Math.PI);
|
---|
| 111 | }
|
---|
| 112 | //taken from https://www.johndcook.com/blog/2009/01/19/stand-alone-error-function-erf/
|
---|
| 113 | private static double StandardNormalDistribution(double x) {
|
---|
| 114 | if (x > 10) return 1;
|
---|
| 115 | if (x < -10) return 0;
|
---|
| 116 | const double a1 = 0.254829592;
|
---|
| 117 | const double a2 = -0.284496736;
|
---|
| 118 | const double a3 = 1.421413741;
|
---|
| 119 | const double a4 = -1.453152027;
|
---|
| 120 | const double a5 = 1.061405429;
|
---|
| 121 | const double p = 0.3275911;
|
---|
| 122 | var sign = x < 0 ? -1 : 1;
|
---|
| 123 | x = Math.Abs(x) / Math.Sqrt(2.0);
|
---|
| 124 | var t = 1.0 / (1.0 + p * x);
|
---|
| 125 | var y = 1.0 - ((((a5 * t + a4) * t + a3) * t + a2) * t + a1) * t * Math.Exp(-x * x);
|
---|
| 126 | return 0.5 * (1.0 + sign * y);
|
---|
| 127 | }
|
---|
| 128 | #endregion
|
---|
| 129 | }
|
---|
| 130 | }
|
---|