[18061] | 1 | using System;
|
---|
| 2 | using System.Collections.Generic;
|
---|
| 3 | using System.Linq;
|
---|
| 4 | using System.Text;
|
---|
| 5 | using System.Threading.Tasks;
|
---|
| 6 | using HeuristicLab.Core;
|
---|
| 7 | using HeuristicLab.Optimization;
|
---|
| 8 | using HEAL.Attic;
|
---|
| 9 | using HeuristicLab.Common;
|
---|
| 10 | using HeuristicLab.Problems.Instances;
|
---|
| 11 | using HeuristicLab.Parameters;
|
---|
| 12 | using HeuristicLab.Data;
|
---|
[18062] | 13 | using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
|
---|
[18075] | 14 | using HeuristicLab.PluginInfrastructure;
|
---|
[18084] | 15 | using HeuristicLab.Problems.Instances.DataAnalysis;
|
---|
[18061] | 16 |
|
---|
[18063] | 17 | namespace HeuristicLab.Problems.DataAnalysis.Symbolic.Regression {
|
---|
[18061] | 18 | [StorableType("7464E84B-65CC-440A-91F0-9FA920D730F9")]
|
---|
[18063] | 19 | [Item(Name = "Structured Symbolic Regression Single Objective Problem (single-objective)", Description = "A problem with a structural definition and unfixed subfunctions.")]
|
---|
| 20 | [Creatable(CreatableAttribute.Categories.GeneticProgrammingProblems, Priority = 150)]
|
---|
[18075] | 21 | public class StructuredSymbolicRegressionSingleObjectiveProblem : SingleObjectiveBasicProblem<MultiEncoding>, IRegressionProblem, IProblemInstanceConsumer<IRegressionProblemData> {
|
---|
[18061] | 22 |
|
---|
| 23 | #region Constants
|
---|
[18081] | 24 | private const string TreeEvaluatorParameterName = "TreeEvaluator";
|
---|
[18061] | 25 | private const string ProblemDataParameterName = "ProblemData";
|
---|
[18063] | 26 | private const string StructureTemplateParameterName = "Structure Template";
|
---|
[18075] | 27 | private const string InterpreterParameterName = "Interpreter";
|
---|
[18076] | 28 | private const string EstimationLimitsParameterName = "EstimationLimits";
|
---|
| 29 | private const string BestTrainingSolutionParameterName = "Best Training Solution";
|
---|
[18072] | 30 |
|
---|
[18076] | 31 | private const string SymbolicExpressionTreeName = "SymbolicExpressionTree";
|
---|
[18095] | 32 | private const string VariableName = "Variable";
|
---|
[18076] | 33 |
|
---|
| 34 | private const string StructureTemplateDescriptionText =
|
---|
[18072] | 35 | "Enter your expression as string in infix format into the empty input field.\n" +
|
---|
| 36 | "By checking the \"Apply Linear Scaling\" checkbox you can add the relevant scaling terms to your expression.\n" +
|
---|
| 37 | "After entering the expression click parse to build the tree.\n" +
|
---|
| 38 | "To edit the defined sub-functions, click on the coressponding colored node in the tree view.";
|
---|
[18061] | 39 | #endregion
|
---|
| 40 |
|
---|
[18072] | 41 | #region Parameters
|
---|
[18081] | 42 | public IConstrainedValueParameter<SymbolicRegressionSingleObjectiveEvaluator> TreeEvaluatorParameter => (IConstrainedValueParameter<SymbolicRegressionSingleObjectiveEvaluator>)Parameters[TreeEvaluatorParameterName];
|
---|
[18061] | 43 | public IValueParameter<IRegressionProblemData> ProblemDataParameter => (IValueParameter<IRegressionProblemData>)Parameters[ProblemDataParameterName];
|
---|
[18063] | 44 | public IFixedValueParameter<StructureTemplate> StructureTemplateParameter => (IFixedValueParameter<StructureTemplate>)Parameters[StructureTemplateParameterName];
|
---|
[18075] | 45 | public IValueParameter<ISymbolicDataAnalysisExpressionTreeInterpreter> InterpreterParameter => (IValueParameter<ISymbolicDataAnalysisExpressionTreeInterpreter>)Parameters[InterpreterParameterName];
|
---|
[18076] | 46 | public IFixedValueParameter<DoubleLimit> EstimationLimitsParameter => (IFixedValueParameter<DoubleLimit>)Parameters[EstimationLimitsParameterName];
|
---|
| 47 | public IResultParameter<ISymbolicRegressionSolution> BestTrainingSolutionParameter => (IResultParameter<ISymbolicRegressionSolution>)Parameters[BestTrainingSolutionParameterName];
|
---|
[18061] | 48 | #endregion
|
---|
| 49 |
|
---|
| 50 | #region Properties
|
---|
[18081] | 51 |
|
---|
[18076] | 52 | public IRegressionProblemData ProblemData {
|
---|
| 53 | get => ProblemDataParameter.Value;
|
---|
[18061] | 54 | set {
|
---|
| 55 | ProblemDataParameter.Value = value;
|
---|
| 56 | ProblemDataChanged?.Invoke(this, EventArgs.Empty);
|
---|
| 57 | }
|
---|
| 58 | }
|
---|
| 59 |
|
---|
[18075] | 60 | public StructureTemplate StructureTemplate => StructureTemplateParameter.Value;
|
---|
[18061] | 61 |
|
---|
[18075] | 62 | public ISymbolicDataAnalysisExpressionTreeInterpreter Interpreter => InterpreterParameter.Value;
|
---|
[18063] | 63 |
|
---|
[18061] | 64 | IParameter IDataAnalysisProblem.ProblemDataParameter => ProblemDataParameter;
|
---|
| 65 | IDataAnalysisProblemData IDataAnalysisProblem.ProblemData => ProblemData;
|
---|
| 66 |
|
---|
[18076] | 67 | public DoubleLimit EstimationLimits => EstimationLimitsParameter.Value;
|
---|
| 68 |
|
---|
[18081] | 69 | public override bool Maximization => false;
|
---|
[18061] | 70 | #endregion
|
---|
| 71 |
|
---|
| 72 | #region EventHandlers
|
---|
| 73 | public event EventHandler ProblemDataChanged;
|
---|
| 74 | #endregion
|
---|
| 75 |
|
---|
| 76 | #region Constructors & Cloning
|
---|
| 77 | public StructuredSymbolicRegressionSingleObjectiveProblem() {
|
---|
[18101] | 78 | var provider = new PhysicsInstanceProvider();
|
---|
| 79 | var descriptor = new SheetBendingProcess();
|
---|
[18084] | 80 | var problemData = provider.LoadData(descriptor);
|
---|
| 81 | var shapeConstraintProblemData = new ShapeConstrainedRegressionProblemData(problemData);
|
---|
| 82 |
|
---|
| 83 |
|
---|
| 84 | var targetInterval = shapeConstraintProblemData.VariableRanges.GetInterval(shapeConstraintProblemData.TargetVariable);
|
---|
[18076] | 85 | var estimationWidth = targetInterval.Width * 10;
|
---|
[18062] | 86 |
|
---|
[18076] | 87 |
|
---|
[18065] | 88 | var structureTemplate = new StructureTemplate();
|
---|
[18066] | 89 | structureTemplate.Changed += OnTemplateChanged;
|
---|
[18065] | 90 |
|
---|
[18081] | 91 | var evaluators = new ItemSet<SymbolicRegressionSingleObjectiveEvaluator>(
|
---|
| 92 | ApplicationManager.Manager.GetInstances<SymbolicRegressionSingleObjectiveEvaluator>()
|
---|
| 93 | .Where(x => x.Maximization == Maximization));
|
---|
| 94 |
|
---|
| 95 | Parameters.Add(new ConstrainedValueParameter<SymbolicRegressionSingleObjectiveEvaluator>(
|
---|
[18095] | 96 | TreeEvaluatorParameterName,
|
---|
[18081] | 97 | evaluators,
|
---|
| 98 | evaluators.First()));
|
---|
| 99 |
|
---|
[18075] | 100 | Parameters.Add(new ValueParameter<IRegressionProblemData>(
|
---|
[18076] | 101 | ProblemDataParameterName,
|
---|
[18084] | 102 | shapeConstraintProblemData));
|
---|
[18081] | 103 | ProblemDataParameter.ValueChanged += ProblemDataParameterValueChanged;
|
---|
| 104 |
|
---|
[18075] | 105 | Parameters.Add(new FixedValueParameter<StructureTemplate>(
|
---|
[18076] | 106 | StructureTemplateParameterName,
|
---|
| 107 | StructureTemplateDescriptionText,
|
---|
[18075] | 108 | structureTemplate));
|
---|
[18099] | 109 |
|
---|
[18075] | 110 | Parameters.Add(new ValueParameter<ISymbolicDataAnalysisExpressionTreeInterpreter>(
|
---|
[18076] | 111 | InterpreterParameterName,
|
---|
| 112 | new SymbolicDataAnalysisExpressionTreeInterpreter()) { Hidden = true });
|
---|
[18099] | 113 |
|
---|
[18076] | 114 | Parameters.Add(new FixedValueParameter<DoubleLimit>(
|
---|
| 115 | EstimationLimitsParameterName,
|
---|
[18095] | 116 | new DoubleLimit(targetInterval.LowerBound - estimationWidth, targetInterval.UpperBound + estimationWidth)) { Hidden = true });
|
---|
[18081] | 117 |
|
---|
[18095] | 118 | Parameters.Add(new ResultParameter<ISymbolicRegressionSolution>(BestTrainingSolutionParameterName, "") { Hidden = true });
|
---|
[18075] | 119 |
|
---|
[18081] | 120 | this.EvaluatorParameter.Hidden = true;
|
---|
[18099] | 121 |
|
---|
[18076] | 122 | Operators.Add(new SymbolicDataAnalysisVariableFrequencyAnalyzer());
|
---|
| 123 | Operators.Add(new MinAverageMaxSymbolicExpressionTreeLengthAnalyzer());
|
---|
| 124 | Operators.Add(new SymbolicExpressionSymbolFrequencyAnalyzer());
|
---|
| 125 |
|
---|
[18099] | 126 | StructureTemplate.Template =
|
---|
[18084] | 127 | "(" +
|
---|
| 128 | "(210000 / (210000 + h)) * ((sigma_y * t * t) / (wR * Rt * t)) + " +
|
---|
| 129 | "PlasticHardening(_) - Elasticity(_)" +
|
---|
| 130 | ")" +
|
---|
| 131 | " * C(_)";
|
---|
[18061] | 132 | }
|
---|
| 133 |
|
---|
[18076] | 134 | public StructuredSymbolicRegressionSingleObjectiveProblem(StructuredSymbolicRegressionSingleObjectiveProblem original,
|
---|
| 135 | Cloner cloner) : base(original, cloner) { }
|
---|
[18061] | 136 |
|
---|
| 137 | [StorableConstructor]
|
---|
[18063] | 138 | protected StructuredSymbolicRegressionSingleObjectiveProblem(StorableConstructorFlag _) : base(_) { }
|
---|
[18065] | 139 | #endregion
|
---|
[18061] | 140 |
|
---|
[18065] | 141 | #region Cloning
|
---|
[18061] | 142 | public override IDeepCloneable Clone(Cloner cloner) =>
|
---|
| 143 | new StructuredSymbolicRegressionSingleObjectiveProblem(this, cloner);
|
---|
| 144 | #endregion
|
---|
| 145 |
|
---|
[18075] | 146 | private void ProblemDataParameterValueChanged(object sender, EventArgs e) {
|
---|
| 147 | StructureTemplate.Reset();
|
---|
| 148 | // InfoBox for Reset?
|
---|
| 149 | }
|
---|
| 150 |
|
---|
[18066] | 151 | private void OnTemplateChanged(object sender, EventArgs args) {
|
---|
[18068] | 152 | SetupStructureTemplate();
|
---|
| 153 | }
|
---|
| 154 |
|
---|
| 155 | private void SetupStructureTemplate() {
|
---|
[18066] | 156 | foreach (var e in Encoding.Encodings.ToArray())
|
---|
| 157 | Encoding.Remove(e);
|
---|
| 158 |
|
---|
[18068] | 159 | foreach (var f in StructureTemplate.SubFunctions.Values) {
|
---|
| 160 | SetupVariables(f);
|
---|
[18075] | 161 | if (!Encoding.Encodings.Any(x => x.Name == f.Name)) // to prevent the same encoding twice
|
---|
| 162 | Encoding.Add(new SymbolicExpressionTreeEncoding(
|
---|
[18076] | 163 | f.Name,
|
---|
| 164 | f.Grammar,
|
---|
| 165 | f.MaximumSymbolicExpressionTreeLength,
|
---|
[18075] | 166 | f.MaximumSymbolicExpressionTreeDepth));
|
---|
[18066] | 167 | }
|
---|
| 168 | }
|
---|
| 169 |
|
---|
| 170 | public override void Analyze(Individual[] individuals, double[] qualities, ResultCollection results, IRandom random) {
|
---|
| 171 | base.Analyze(individuals, qualities, results, random);
|
---|
| 172 |
|
---|
[18095] | 173 | var best = GetBestIndividual(individuals, qualities).Item1;
|
---|
[18076] | 174 |
|
---|
| 175 | if (!results.ContainsKey(BestTrainingSolutionParameter.ActualName)) {
|
---|
| 176 | results.Add(new Result(BestTrainingSolutionParameter.ActualName, typeof(SymbolicRegressionSolution)));
|
---|
[18066] | 177 | }
|
---|
| 178 |
|
---|
[18076] | 179 | var tree = (ISymbolicExpressionTree)best[SymbolicExpressionTreeName];
|
---|
| 180 |
|
---|
| 181 | var model = new SymbolicRegressionModel(ProblemData.TargetVariable, tree, Interpreter);
|
---|
| 182 | var solution = model.CreateRegressionSolution(ProblemData);
|
---|
| 183 |
|
---|
| 184 | results[BestTrainingSolutionParameter.ActualName].Value = solution;
|
---|
[18066] | 185 | }
|
---|
| 186 |
|
---|
[18076] | 187 |
|
---|
[18065] | 188 | public override double Evaluate(Individual individual, IRandom random) {
|
---|
[18066] | 189 | var tree = BuildTree(individual);
|
---|
[18071] | 190 |
|
---|
[18072] | 191 | if (StructureTemplate.ApplyLinearScaling)
|
---|
[18076] | 192 | AdjustLinearScalingParams(ProblemData, tree, Interpreter);
|
---|
| 193 |
|
---|
| 194 | individual[SymbolicExpressionTreeName] = tree;
|
---|
| 195 |
|
---|
[18103] | 196 | if(TreeEvaluatorParameter.Value is SymbolicRegressionConstantOptimizationEvaluator)
|
---|
| 197 | TreeEvaluatorParameter.Value.RandomParameter.Value = random;
|
---|
| 198 |
|
---|
[18095] | 199 | return TreeEvaluatorParameter.Value.Evaluate(
|
---|
[18103] | 200 | tree, ProblemData,
|
---|
| 201 | ProblemData.TrainingIndices,
|
---|
[18095] | 202 | Interpreter,
|
---|
| 203 | StructureTemplate.ApplyLinearScaling,
|
---|
[18099] | 204 | EstimationLimits.Lower,
|
---|
[18095] | 205 | EstimationLimits.Upper);
|
---|
[18066] | 206 | }
|
---|
| 207 |
|
---|
[18076] | 208 | private static void AdjustLinearScalingParams(IRegressionProblemData problemData, ISymbolicExpressionTree tree, ISymbolicDataAnalysisExpressionTreeInterpreter interpreter) {
|
---|
[18071] | 209 | var offsetNode = tree.Root.GetSubtree(0).GetSubtree(0);
|
---|
| 210 | var scalingNode = offsetNode.Subtrees.Where(x => !(x is ConstantTreeNode)).First();
|
---|
| 211 |
|
---|
| 212 | var offsetConstantNode = (ConstantTreeNode)offsetNode.Subtrees.Where(x => x is ConstantTreeNode).First();
|
---|
| 213 | var scalingConstantNode = (ConstantTreeNode)scalingNode.Subtrees.Where(x => x is ConstantTreeNode).First();
|
---|
| 214 |
|
---|
[18076] | 215 | var estimatedValues = interpreter.GetSymbolicExpressionTreeValues(tree, problemData.Dataset, problemData.TrainingIndices);
|
---|
| 216 | var targetValues = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, problemData.TrainingIndices);
|
---|
[18071] | 217 |
|
---|
| 218 | OnlineLinearScalingParameterCalculator.Calculate(estimatedValues, targetValues, out double a, out double b, out OnlineCalculatorError error);
|
---|
[18076] | 219 | if (error == OnlineCalculatorError.None) {
|
---|
[18071] | 220 | offsetConstantNode.Value = a;
|
---|
| 221 | scalingConstantNode.Value = b;
|
---|
| 222 | }
|
---|
| 223 | }
|
---|
| 224 |
|
---|
[18066] | 225 | private ISymbolicExpressionTree BuildTree(Individual individual) {
|
---|
[18075] | 226 | if (StructureTemplate.Tree == null)
|
---|
| 227 | throw new ArgumentException("No structure template defined!");
|
---|
| 228 |
|
---|
[18065] | 229 | var templateTree = (ISymbolicExpressionTree)StructureTemplate.Tree.Clone();
|
---|
[18066] | 230 |
|
---|
[18065] | 231 | // build main tree
|
---|
[18076] | 232 | foreach (var subFunctionTreeNode in templateTree.IterateNodesPrefix().OfType<SubFunctionTreeNode>()) {
|
---|
| 233 | var subFunctionTree = individual.SymbolicExpressionTree(subFunctionTreeNode.Name);
|
---|
[18062] | 234 |
|
---|
[18076] | 235 | // add new tree
|
---|
| 236 | var subTree = subFunctionTree.Root.GetSubtree(0) // Start
|
---|
| 237 | .GetSubtree(0); // Offset
|
---|
| 238 | subTree = (ISymbolicExpressionTreeNode)subTree.Clone();
|
---|
| 239 | subFunctionTreeNode.AddSubtree(subTree);
|
---|
| 240 |
|
---|
[18065] | 241 | }
|
---|
[18066] | 242 | return templateTree;
|
---|
[18061] | 243 | }
|
---|
| 244 |
|
---|
[18068] | 245 | private void SetupVariables(SubFunction subFunction) {
|
---|
[18095] | 246 | var varSym = (Variable)subFunction.Grammar.GetSymbol(VariableName);
|
---|
[18068] | 247 | if (varSym == null) {
|
---|
| 248 | varSym = new Variable();
|
---|
| 249 | subFunction.Grammar.AddSymbol(varSym);
|
---|
| 250 | }
|
---|
| 251 |
|
---|
| 252 | var allVariables = ProblemData.InputVariables.Select(x => x.Value);
|
---|
| 253 | var allInputs = allVariables.Where(x => x != ProblemData.TargetVariable);
|
---|
| 254 |
|
---|
| 255 | // set all variables
|
---|
| 256 | varSym.AllVariableNames = allVariables;
|
---|
| 257 |
|
---|
| 258 | // set all allowed variables
|
---|
| 259 | if (subFunction.Arguments.Contains("_")) {
|
---|
| 260 | varSym.VariableNames = allInputs;
|
---|
| 261 | } else {
|
---|
| 262 | var vars = new List<string>();
|
---|
| 263 | var exceptions = new List<Exception>();
|
---|
| 264 | foreach (var arg in subFunction.Arguments) {
|
---|
| 265 | if (allInputs.Contains(arg))
|
---|
| 266 | vars.Add(arg);
|
---|
| 267 | else
|
---|
| 268 | exceptions.Add(new ArgumentException($"The argument '{arg}' for sub-function '{subFunction.Name}' is not a valid variable."));
|
---|
| 269 | }
|
---|
| 270 | if (exceptions.Any())
|
---|
| 271 | throw new AggregateException(exceptions);
|
---|
| 272 | varSym.VariableNames = vars;
|
---|
| 273 | }
|
---|
| 274 |
|
---|
| 275 | varSym.Enabled = true;
|
---|
| 276 | }
|
---|
| 277 |
|
---|
[18099] | 278 | public void Load(IRegressionProblemData data) {
|
---|
| 279 | ProblemData = data;
|
---|
| 280 | StructureTemplate.Template = "f(_)";
|
---|
| 281 | }
|
---|
[18061] | 282 | }
|
---|
| 283 | }
|
---|