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;
|
---|
13 | using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
|
---|
14 |
|
---|
15 | namespace HeuristicLab.Problems.DataAnalysis.Symbolic.Regression {
|
---|
16 | [StorableType("7464E84B-65CC-440A-91F0-9FA920D730F9")]
|
---|
17 | [Item(Name = "Structured Symbolic Regression Single Objective Problem (single-objective)", Description = "A problem with a structural definition and unfixed subfunctions.")]
|
---|
18 | [Creatable(CreatableAttribute.Categories.GeneticProgrammingProblems, Priority = 150)]
|
---|
19 | public class StructuredSymbolicRegressionSingleObjectiveProblem : SingleObjectiveBasicProblem<MultiEncoding>, IRegressionProblem, IProblemInstanceConsumer<RegressionProblemData> {
|
---|
20 |
|
---|
21 | #region Constants
|
---|
22 | private const string ProblemDataParameterName = "ProblemData";
|
---|
23 | private const string StructureDefinitionParameterName = "Structure Definition";
|
---|
24 | private const string StructureTemplateParameterName = "Structure Template";
|
---|
25 |
|
---|
26 | private const string StructureTemplateDescriptionText =
|
---|
27 | "Enter your expression as string in infix format into the empty input field.\n" +
|
---|
28 | "By checking the \"Apply Linear Scaling\" checkbox you can add the relevant scaling terms to your expression.\n" +
|
---|
29 | "After entering the expression click parse to build the tree.\n" +
|
---|
30 | "To edit the defined sub-functions, click on the coressponding colored node in the tree view.";
|
---|
31 | #endregion
|
---|
32 |
|
---|
33 | #region Parameters
|
---|
34 | public IValueParameter<IRegressionProblemData> ProblemDataParameter => (IValueParameter<IRegressionProblemData>)Parameters[ProblemDataParameterName];
|
---|
35 | public IFixedValueParameter<StringValue> StructureDefinitionParameter => (IFixedValueParameter<StringValue>)Parameters[StructureDefinitionParameterName];
|
---|
36 | public IFixedValueParameter<StructureTemplate> StructureTemplateParameter => (IFixedValueParameter<StructureTemplate>)Parameters[StructureTemplateParameterName];
|
---|
37 | #endregion
|
---|
38 |
|
---|
39 | #region Properties
|
---|
40 | public IRegressionProblemData ProblemData {
|
---|
41 | get => ProblemDataParameter.Value;
|
---|
42 | set {
|
---|
43 | ProblemDataParameter.Value = value;
|
---|
44 | ProblemDataChanged?.Invoke(this, EventArgs.Empty);
|
---|
45 | }
|
---|
46 | }
|
---|
47 |
|
---|
48 | public string StructureDefinition {
|
---|
49 | get => StructureDefinitionParameter.Value.Value;
|
---|
50 | set => StructureDefinitionParameter.Value.Value = value;
|
---|
51 | }
|
---|
52 |
|
---|
53 | public StructureTemplate StructureTemplate {
|
---|
54 | get => StructureTemplateParameter.Value;
|
---|
55 | }
|
---|
56 |
|
---|
57 | public ISymbolicDataAnalysisExpressionTreeInterpreter Interpreter { get; } = new SymbolicDataAnalysisExpressionTreeInterpreter();
|
---|
58 |
|
---|
59 | IParameter IDataAnalysisProblem.ProblemDataParameter => ProblemDataParameter;
|
---|
60 | IDataAnalysisProblemData IDataAnalysisProblem.ProblemData => ProblemData;
|
---|
61 |
|
---|
62 | public override bool Maximization => true;
|
---|
63 | #endregion
|
---|
64 |
|
---|
65 | #region EventHandlers
|
---|
66 | public event EventHandler ProblemDataChanged;
|
---|
67 | #endregion
|
---|
68 |
|
---|
69 | #region Constructors & Cloning
|
---|
70 | public StructuredSymbolicRegressionSingleObjectiveProblem() {
|
---|
71 | var problemData = new ShapeConstrainedRegressionProblemData();
|
---|
72 |
|
---|
73 | var structureTemplate = new StructureTemplate();
|
---|
74 | structureTemplate.Changed += OnTemplateChanged;
|
---|
75 |
|
---|
76 | Parameters.Add(new ValueParameter<IRegressionProblemData>(ProblemDataParameterName, problemData));
|
---|
77 | Parameters.Add(new FixedValueParameter<StructureTemplate>(StructureTemplateParameterName,
|
---|
78 | StructureTemplateDescriptionText, structureTemplate));
|
---|
79 |
|
---|
80 |
|
---|
81 | }
|
---|
82 |
|
---|
83 | public StructuredSymbolicRegressionSingleObjectiveProblem(StructuredSymbolicRegressionSingleObjectiveProblem original,
|
---|
84 | Cloner cloner) : base(original, cloner){ }
|
---|
85 |
|
---|
86 | [StorableConstructor]
|
---|
87 | protected StructuredSymbolicRegressionSingleObjectiveProblem(StorableConstructorFlag _) : base(_) { }
|
---|
88 | #endregion
|
---|
89 |
|
---|
90 | #region Cloning
|
---|
91 | public override IDeepCloneable Clone(Cloner cloner) =>
|
---|
92 | new StructuredSymbolicRegressionSingleObjectiveProblem(this, cloner);
|
---|
93 | #endregion
|
---|
94 |
|
---|
95 | private void OnTemplateChanged(object sender, EventArgs args) {
|
---|
96 | SetupStructureTemplate();
|
---|
97 | }
|
---|
98 |
|
---|
99 | private void SetupStructureTemplate() {
|
---|
100 | foreach (var e in Encoding.Encodings.ToArray())
|
---|
101 | Encoding.Remove(e);
|
---|
102 |
|
---|
103 | foreach (var f in StructureTemplate.SubFunctions.Values) {
|
---|
104 | SetupVariables(f);
|
---|
105 | if(!Encoding.Encodings.Any(x => x.Name == f.Name)) // to prevent the same encoding twice
|
---|
106 | Encoding.Add(new SymbolicExpressionTreeEncoding(f.Name, f.Grammar, f.MaximumSymbolicExpressionTreeLength, f.MaximumSymbolicExpressionTreeDepth));
|
---|
107 | }
|
---|
108 | }
|
---|
109 |
|
---|
110 | public override void Analyze(Individual[] individuals, double[] qualities, ResultCollection results, IRandom random) {
|
---|
111 | base.Analyze(individuals, qualities, results, random);
|
---|
112 |
|
---|
113 | int bestIdx = 0;
|
---|
114 | double bestQuality = Maximization ? double.MinValue : double.MaxValue;
|
---|
115 | for(int idx = 0; idx < qualities.Length; ++idx) {
|
---|
116 | if((Maximization && qualities[idx] > bestQuality) ||
|
---|
117 | (!Maximization && qualities[idx] < bestQuality)) {
|
---|
118 | bestQuality = qualities[idx];
|
---|
119 | bestIdx = idx;
|
---|
120 | }
|
---|
121 | }
|
---|
122 |
|
---|
123 | if (results.TryGetValue("Best Tree", out IResult result)) {
|
---|
124 | var tree = BuildTree(individuals[bestIdx]);
|
---|
125 | if (StructureTemplate.ApplyLinearScaling)
|
---|
126 | AdjustLinearScalingParams(tree, Interpreter);
|
---|
127 | result.Value = tree;
|
---|
128 | }
|
---|
129 | else {
|
---|
130 | var tree = BuildTree(individuals[bestIdx]);
|
---|
131 | if (StructureTemplate.ApplyLinearScaling)
|
---|
132 | AdjustLinearScalingParams(tree, Interpreter);
|
---|
133 | results.Add(new Result("Best Tree", tree));
|
---|
134 | }
|
---|
135 |
|
---|
136 | }
|
---|
137 |
|
---|
138 | public override double Evaluate(Individual individual, IRandom random) {
|
---|
139 | var tree = BuildTree(individual);
|
---|
140 |
|
---|
141 | if (StructureTemplate.ApplyLinearScaling)
|
---|
142 | AdjustLinearScalingParams(tree, Interpreter);
|
---|
143 | var estimationInterval = ProblemData.VariableRanges.GetInterval(ProblemData.TargetVariable);
|
---|
144 | var quality = SymbolicRegressionSingleObjectivePearsonRSquaredEvaluator.Calculate(
|
---|
145 | Interpreter, tree,
|
---|
146 | estimationInterval.LowerBound, estimationInterval.UpperBound,
|
---|
147 | ProblemData, ProblemData.TrainingIndices, false);
|
---|
148 |
|
---|
149 | return quality;
|
---|
150 | }
|
---|
151 |
|
---|
152 | private void AdjustLinearScalingParams(ISymbolicExpressionTree tree, ISymbolicDataAnalysisExpressionTreeInterpreter interpreter) {
|
---|
153 | var offsetNode = tree.Root.GetSubtree(0).GetSubtree(0);
|
---|
154 | var scalingNode = offsetNode.Subtrees.Where(x => !(x is ConstantTreeNode)).First();
|
---|
155 |
|
---|
156 | var offsetConstantNode = (ConstantTreeNode)offsetNode.Subtrees.Where(x => x is ConstantTreeNode).First();
|
---|
157 | var scalingConstantNode = (ConstantTreeNode)scalingNode.Subtrees.Where(x => x is ConstantTreeNode).First();
|
---|
158 |
|
---|
159 | var estimatedValues = interpreter.GetSymbolicExpressionTreeValues(tree, ProblemData.Dataset, ProblemData.TrainingIndices);
|
---|
160 | var targetValues = ProblemData.Dataset.GetDoubleValues(ProblemData.TargetVariable, ProblemData.TrainingIndices);
|
---|
161 |
|
---|
162 | OnlineLinearScalingParameterCalculator.Calculate(estimatedValues, targetValues, out double a, out double b, out OnlineCalculatorError error);
|
---|
163 | if(error == OnlineCalculatorError.None) {
|
---|
164 | offsetConstantNode.Value = a;
|
---|
165 | scalingConstantNode.Value = b;
|
---|
166 | }
|
---|
167 | }
|
---|
168 |
|
---|
169 | private ISymbolicExpressionTree BuildTree(Individual individual) {
|
---|
170 | var templateTree = (ISymbolicExpressionTree)StructureTemplate.Tree.Clone();
|
---|
171 |
|
---|
172 | // build main tree
|
---|
173 | foreach (var n in templateTree.IterateNodesPrefix()) {
|
---|
174 | if (n.Symbol is SubFunctionSymbol) {
|
---|
175 | var subFunctionTreeNode = n as SubFunctionTreeNode;
|
---|
176 | var subFunctionTree = individual.SymbolicExpressionTree(subFunctionTreeNode.Name);
|
---|
177 | //var parent = n.Parent;
|
---|
178 |
|
---|
179 | // remove SubFunctionTreeNode
|
---|
180 | //parent.RemoveSubtree(parent.IndexOfSubtree(subFunctionTreeNode));
|
---|
181 |
|
---|
182 | // add new tree
|
---|
183 | var subTree = subFunctionTree.Root.GetSubtree(0) // Start
|
---|
184 | .GetSubtree(0); // Offset
|
---|
185 | //parent.AddSubtree(subTree);
|
---|
186 | subFunctionTreeNode.AddSubtree(subTree);
|
---|
187 | }
|
---|
188 | }
|
---|
189 | return templateTree;
|
---|
190 | }
|
---|
191 |
|
---|
192 | private void SetupVariables(SubFunction subFunction) {
|
---|
193 | var varSym = (Variable)subFunction.Grammar.GetSymbol("Variable");
|
---|
194 | if (varSym == null) {
|
---|
195 | varSym = new Variable();
|
---|
196 | subFunction.Grammar.AddSymbol(varSym);
|
---|
197 | }
|
---|
198 |
|
---|
199 | var allVariables = ProblemData.InputVariables.Select(x => x.Value);
|
---|
200 | var allInputs = allVariables.Where(x => x != ProblemData.TargetVariable);
|
---|
201 |
|
---|
202 | // set all variables
|
---|
203 | varSym.AllVariableNames = allVariables;
|
---|
204 |
|
---|
205 | // set all allowed variables
|
---|
206 | if (subFunction.Arguments.Contains("_")) {
|
---|
207 | varSym.VariableNames = allInputs;
|
---|
208 | } else {
|
---|
209 | var vars = new List<string>();
|
---|
210 | var exceptions = new List<Exception>();
|
---|
211 | foreach (var arg in subFunction.Arguments) {
|
---|
212 | if (allInputs.Contains(arg))
|
---|
213 | vars.Add(arg);
|
---|
214 | else
|
---|
215 | exceptions.Add(new ArgumentException($"The argument '{arg}' for sub-function '{subFunction.Name}' is not a valid variable."));
|
---|
216 | }
|
---|
217 | if (exceptions.Any())
|
---|
218 | throw new AggregateException(exceptions);
|
---|
219 | varSym.VariableNames = vars;
|
---|
220 | }
|
---|
221 |
|
---|
222 | varSym.Enabled = true;
|
---|
223 | }
|
---|
224 |
|
---|
225 | public void Load(RegressionProblemData data) {
|
---|
226 | ProblemData = data;
|
---|
227 | SetupStructureTemplate();
|
---|
228 | }
|
---|
229 | }
|
---|
230 | }
|
---|