[17958] | 1 | #region License Information
|
---|
| 2 |
|
---|
| 3 | /* HeuristicLab
|
---|
| 4 | * Copyright (C) Heuristic and Evolutionary Algorithms Laboratory (HEAL)
|
---|
| 5 | *
|
---|
| 6 | * This file is part of HeuristicLab.
|
---|
| 7 | *
|
---|
| 8 | * HeuristicLab is free software: you can redistribute it and/or modify
|
---|
| 9 | * it under the terms of the GNU General Public License as published by
|
---|
| 10 | * the Free Software Foundation, either version 3 of the License, or
|
---|
| 11 | * (at your option) any later version.
|
---|
| 12 | *
|
---|
| 13 | * HeuristicLab is distributed in the hope that it will be useful,
|
---|
| 14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
---|
| 15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
---|
| 16 | * GNU General Public License for more details.
|
---|
| 17 | *
|
---|
| 18 | * You should have received a copy of the GNU General Public License
|
---|
| 19 | * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.
|
---|
| 20 | */
|
---|
| 21 |
|
---|
| 22 | #endregion
|
---|
| 23 |
|
---|
| 24 | using System;
|
---|
| 25 | using System.Collections.Generic;
|
---|
| 26 | using System.Linq;
|
---|
| 27 | using HEAL.Attic;
|
---|
| 28 | using HeuristicLab.Common;
|
---|
| 29 | using HeuristicLab.Core;
|
---|
| 30 | using HeuristicLab.Data;
|
---|
| 31 | using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
|
---|
| 32 | using HeuristicLab.Parameters;
|
---|
| 33 |
|
---|
| 34 | namespace HeuristicLab.Problems.DataAnalysis.Symbolic.Regression {
|
---|
| 35 | [Item("NMSE Evaluator with shape constraints (multi-objective)",
|
---|
| 36 | "Calculates the NMSE and constraint violations for a symbolic regression model.")]
|
---|
| 37 | [StorableType("8E9D76B7-ED9C-43E7-9898-01FBD3633880")]
|
---|
| 38 | public class NMSEMultiObjectiveConstraintsEvaluator : SymbolicRegressionMultiObjectiveEvaluator, IMultiObjectiveConstraintsEvaluator {
|
---|
| 39 | private const string NumConstraintsParameterName = "NumConstraints";
|
---|
| 40 | private const string BoundsEstimatorParameterName = "BoundsEstimator";
|
---|
| 41 |
|
---|
| 42 | public IFixedValueParameter<IntValue> NumConstraintsParameter =>
|
---|
| 43 | (IFixedValueParameter<IntValue>)Parameters[NumConstraintsParameterName];
|
---|
| 44 |
|
---|
| 45 | public IValueParameter<IBoundsEstimator> BoundsEstimatorParameter =>
|
---|
| 46 | (IValueParameter<IBoundsEstimator>)Parameters[BoundsEstimatorParameterName];
|
---|
| 47 |
|
---|
| 48 | public int NumConstraints {
|
---|
| 49 | get => NumConstraintsParameter.Value.Value;
|
---|
| 50 | set {
|
---|
| 51 | NumConstraintsParameter.Value.Value = value;
|
---|
| 52 | }
|
---|
| 53 | }
|
---|
| 54 |
|
---|
| 55 | public IBoundsEstimator BoundsEstimator {
|
---|
| 56 | get => BoundsEstimatorParameter.Value;
|
---|
| 57 | set => BoundsEstimatorParameter.Value = value;
|
---|
| 58 | }
|
---|
| 59 |
|
---|
| 60 | public override IEnumerable<bool> Maximization => new bool[1 + NumConstraints]; // minimize all objectives
|
---|
| 61 |
|
---|
| 62 | #region Constructors
|
---|
| 63 |
|
---|
| 64 | public NMSEMultiObjectiveConstraintsEvaluator() {
|
---|
| 65 | Parameters.Add(new FixedValueParameter<IntValue>(NumConstraintsParameterName, new IntValue(0)));
|
---|
| 66 | Parameters.Add(new ValueParameter<IBoundsEstimator>(BoundsEstimatorParameterName, new IntervalArithBoundsEstimator()));
|
---|
| 67 | }
|
---|
| 68 |
|
---|
| 69 | [StorableConstructor]
|
---|
| 70 | protected NMSEMultiObjectiveConstraintsEvaluator(StorableConstructorFlag _) : base(_) { }
|
---|
| 71 |
|
---|
| 72 | protected NMSEMultiObjectiveConstraintsEvaluator(NMSEMultiObjectiveConstraintsEvaluator original, Cloner cloner) : base(original, cloner) { }
|
---|
| 73 |
|
---|
| 74 | #endregion
|
---|
| 75 |
|
---|
| 76 | [StorableHook(HookType.AfterDeserialization)]
|
---|
| 77 | private void AfterDeserialization() { }
|
---|
| 78 |
|
---|
| 79 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
| 80 | return new NMSEMultiObjectiveConstraintsEvaluator(this, cloner);
|
---|
| 81 | }
|
---|
| 82 |
|
---|
| 83 |
|
---|
| 84 | public override IOperation InstrumentedApply() {
|
---|
| 85 | var rows = GenerateRowsToEvaluate();
|
---|
| 86 | var tree = SymbolicExpressionTreeParameter.ActualValue;
|
---|
| 87 | var problemData = ProblemDataParameter.ActualValue;
|
---|
| 88 | var interpreter = SymbolicDataAnalysisTreeInterpreterParameter.ActualValue;
|
---|
| 89 | var estimationLimits = EstimationLimitsParameter.ActualValue;
|
---|
| 90 | var applyLinearScaling = ApplyLinearScalingParameter.ActualValue.Value;
|
---|
| 91 |
|
---|
| 92 | if (UseConstantOptimization) {
|
---|
| 93 | SymbolicRegressionConstantOptimizationEvaluator.OptimizeConstants(interpreter, tree, problemData, rows,
|
---|
| 94 | false,
|
---|
| 95 | ConstantOptimizationIterations,
|
---|
| 96 | ConstantOptimizationUpdateVariableWeights,
|
---|
| 97 | estimationLimits.Lower,
|
---|
| 98 | estimationLimits.Upper);
|
---|
| 99 | } else {
|
---|
| 100 | if (applyLinearScaling) {
|
---|
| 101 | var rootNode = new ProgramRootSymbol().CreateTreeNode();
|
---|
| 102 | var startNode = new StartSymbol().CreateTreeNode();
|
---|
| 103 | var offset = tree.Root.GetSubtree(0) //Start
|
---|
| 104 | .GetSubtree(0); //Offset
|
---|
| 105 | var scaling = offset.GetSubtree(0);
|
---|
| 106 |
|
---|
| 107 | //Check if tree contains offset and scaling nodes
|
---|
| 108 | if (!(offset.Symbol is Addition) || !(scaling.Symbol is Multiplication))
|
---|
| 109 | throw new ArgumentException($"{ItemName} can only be used with LinearScalingGrammar.");
|
---|
| 110 |
|
---|
| 111 |
|
---|
| 112 | var t = (ISymbolicExpressionTreeNode)scaling.GetSubtree(0).Clone();
|
---|
| 113 | rootNode.AddSubtree(startNode);
|
---|
| 114 | startNode.AddSubtree(t);
|
---|
| 115 | var newTree = new SymbolicExpressionTree(rootNode);
|
---|
| 116 |
|
---|
| 117 | //calculate alpha and beta for scaling
|
---|
| 118 | var estimatedValues = interpreter.GetSymbolicExpressionTreeValues(newTree, problemData.Dataset, rows);
|
---|
| 119 |
|
---|
| 120 | var targetValues = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, rows);
|
---|
| 121 | OnlineLinearScalingParameterCalculator.Calculate(estimatedValues, targetValues, out var alpha, out var beta,
|
---|
| 122 | out var errorState);
|
---|
| 123 | if (errorState == OnlineCalculatorError.None) {
|
---|
| 124 | //Set alpha and beta to the scaling nodes from ia grammar
|
---|
| 125 | var offsetParameter = offset.GetSubtree(1) as ConstantTreeNode;
|
---|
| 126 | offsetParameter.Value = alpha;
|
---|
| 127 | var scalingParameter = scaling.GetSubtree(1) as ConstantTreeNode;
|
---|
| 128 | scalingParameter.Value = beta;
|
---|
| 129 | }
|
---|
| 130 | } // else alpha and beta are evolved
|
---|
| 131 | }
|
---|
| 132 |
|
---|
| 133 | var qualities = Calculate(interpreter, tree, estimationLimits.Lower, estimationLimits.Upper, problemData,
|
---|
| 134 | rows, BoundsEstimator, DecimalPlaces);
|
---|
| 135 | QualitiesParameter.ActualValue = new DoubleArray(qualities);
|
---|
| 136 | return base.InstrumentedApply();
|
---|
| 137 | }
|
---|
| 138 |
|
---|
| 139 | public override double[] Evaluate(
|
---|
| 140 | IExecutionContext context, ISymbolicExpressionTree tree,
|
---|
| 141 | IRegressionProblemData problemData,
|
---|
| 142 | IEnumerable<int> rows) {
|
---|
| 143 | SymbolicDataAnalysisTreeInterpreterParameter.ExecutionContext = context;
|
---|
| 144 | EstimationLimitsParameter.ExecutionContext = context;
|
---|
| 145 | ApplyLinearScalingParameter.ExecutionContext = context;
|
---|
| 146 |
|
---|
| 147 | var quality = Calculate(SymbolicDataAnalysisTreeInterpreterParameter.ActualValue, tree,
|
---|
| 148 | EstimationLimitsParameter.ActualValue.Lower, EstimationLimitsParameter.ActualValue.Upper,
|
---|
| 149 | problemData, rows, BoundsEstimator, DecimalPlaces);
|
---|
| 150 |
|
---|
| 151 | SymbolicDataAnalysisTreeInterpreterParameter.ExecutionContext = null;
|
---|
| 152 | EstimationLimitsParameter.ExecutionContext = null;
|
---|
| 153 | ApplyLinearScalingParameter.ExecutionContext = null;
|
---|
| 154 |
|
---|
| 155 | return quality;
|
---|
| 156 | }
|
---|
| 157 |
|
---|
| 158 |
|
---|
| 159 | public static double[] Calculate(
|
---|
| 160 | ISymbolicDataAnalysisExpressionTreeInterpreter interpreter,
|
---|
| 161 | ISymbolicExpressionTree solution, double lowerEstimationLimit,
|
---|
| 162 | double upperEstimationLimit,
|
---|
| 163 | IRegressionProblemData problemData, IEnumerable<int> rows, IBoundsEstimator estimator, int decimalPlaces) {
|
---|
| 164 | OnlineCalculatorError errorState;
|
---|
| 165 | var estimatedValues = interpreter.GetSymbolicExpressionTreeValues(solution, problemData.Dataset, rows);
|
---|
| 166 | var targetValues = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, rows);
|
---|
| 167 | var constraints = Enumerable.Empty<ShapeConstraint>();
|
---|
| 168 | if (problemData is ShapeConstrainedRegressionProblemData scProbData) {
|
---|
| 169 | constraints = scProbData.ShapeConstraints.EnabledConstraints;
|
---|
| 170 | }
|
---|
| 171 | var intervalCollection = problemData.VariableRanges;
|
---|
| 172 |
|
---|
| 173 | double nmse;
|
---|
| 174 |
|
---|
| 175 | var boundedEstimatedValues = estimatedValues.LimitToRange(lowerEstimationLimit, upperEstimationLimit);
|
---|
| 176 | nmse = OnlineNormalizedMeanSquaredErrorCalculator.Calculate(targetValues, boundedEstimatedValues, out errorState);
|
---|
| 177 |
|
---|
| 178 | if (errorState != OnlineCalculatorError.None) nmse = 1.0;
|
---|
| 179 |
|
---|
| 180 | if (decimalPlaces >= 0)
|
---|
| 181 | nmse = Math.Round(nmse, decimalPlaces);
|
---|
| 182 |
|
---|
| 183 | if (nmse > 1)
|
---|
| 184 | nmse = 1.0;
|
---|
| 185 |
|
---|
| 186 | var objectives = new List<double> { nmse };
|
---|
| 187 | var violations = IntervalUtil.GetConstraintViolations(constraints, estimator, intervalCollection, solution);
|
---|
| 188 | foreach (var violation in violations) {
|
---|
| 189 | if (double.IsNaN(violation) || double.IsInfinity(violation)) {
|
---|
| 190 | objectives.Add(double.MaxValue);
|
---|
| 191 | } else {
|
---|
| 192 | objectives.Add(Math.Round(violation, decimalPlaces));
|
---|
| 193 | }
|
---|
| 194 | }
|
---|
| 195 |
|
---|
| 196 | return objectives.ToArray();
|
---|
| 197 | }
|
---|
| 198 | }
|
---|
| 199 | } |
---|