1 | #region License Information
|
---|
2 |
|
---|
3 | /* HeuristicLab
|
---|
4 | * Copyright (C) 2002-2015 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.Linq;
|
---|
25 | using HeuristicLab.Common;
|
---|
26 | using HeuristicLab.Core;
|
---|
27 | using HeuristicLab.Data;
|
---|
28 | using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
|
---|
29 | using HeuristicLab.Operators;
|
---|
30 | using HeuristicLab.Parameters;
|
---|
31 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
32 |
|
---|
33 | namespace HeuristicLab.Problems.DataAnalysis.Symbolic {
|
---|
34 | [StorableClass]
|
---|
35 | [Item("SymbolicExpressionTreePruningOperator", "An operator that replaces introns with constant values in a symbolic expression tree.")]
|
---|
36 | public abstract class SymbolicDataAnalysisExpressionPruningOperator : SingleSuccessorOperator, ISymbolicExpressionTreeOperator {
|
---|
37 | #region parameter names
|
---|
38 | private const string ProblemDataParameterName = "ProblemData";
|
---|
39 | private const string SymbolicDataAnalysisModelParameterName = "SymbolicDataAnalysisModel";
|
---|
40 | private const string ImpactValuesCalculatorParameterName = "ImpactValuesCalculator";
|
---|
41 | private const string PrunedSubtreesParameterName = "PrunedSubtrees";
|
---|
42 | private const string PrunedTreesParameterName = "PrunedTrees";
|
---|
43 | private const string FitnessCalculationPartitionParameterName = "FitnessCalculationPartition";
|
---|
44 | private const string NodeImpactThresholdParameterName = "ImpactThreshold";
|
---|
45 | private const string PruneOnlyZeroImpactNodesParameterName = "PruneOnlyZeroImpactNodes";
|
---|
46 | private const string SymbolicExpressionTreeParameterName = "SymbolicExpressionTree"; // the tree to be pruned
|
---|
47 | private const string QualityParameterName = "Quality"; // the quality
|
---|
48 | private const string EstimationLimitsParameterName = "EstimationLimits";
|
---|
49 | private const string InterpreterParameterName = "SymbolicExpressionTreeInterpreter";
|
---|
50 | #endregion
|
---|
51 |
|
---|
52 | #region parameter properties
|
---|
53 | public ILookupParameter<ISymbolicExpressionTree> SymbolicExpressionTreeParameter {
|
---|
54 | get { return (ILookupParameter<ISymbolicExpressionTree>)Parameters[SymbolicExpressionTreeParameterName]; }
|
---|
55 | }
|
---|
56 | public ILookupParameter<DoubleValue> QualityParameter {
|
---|
57 | get { return (ILookupParameter<DoubleValue>)Parameters[QualityParameterName]; }
|
---|
58 | }
|
---|
59 | public ILookupParameter<IDataAnalysisProblemData> ProblemDataParameter {
|
---|
60 | get { return (ILookupParameter<IDataAnalysisProblemData>)Parameters[ProblemDataParameterName]; }
|
---|
61 | }
|
---|
62 | public IValueParameter<ISymbolicDataAnalysisSolutionImpactValuesCalculator> ImpactValuesCalculatorParameter {
|
---|
63 | get { return (IValueParameter<ISymbolicDataAnalysisSolutionImpactValuesCalculator>)Parameters[ImpactValuesCalculatorParameterName]; }
|
---|
64 | }
|
---|
65 | public ILookupParameter<IntRange> FitnessCalculationPartitionParameter {
|
---|
66 | get { return (ILookupParameter<IntRange>)Parameters[FitnessCalculationPartitionParameterName]; }
|
---|
67 | }
|
---|
68 | public ILookupParameter<IntValue> PrunedSubtreesParameter {
|
---|
69 | get { return (ILookupParameter<IntValue>)Parameters[PrunedSubtreesParameterName]; }
|
---|
70 | }
|
---|
71 | public ILookupParameter<IntValue> PrunedTreesParameter {
|
---|
72 | get { return (ILookupParameter<IntValue>)Parameters[PrunedTreesParameterName]; }
|
---|
73 | }
|
---|
74 | public IFixedValueParameter<DoubleValue> NodeImpactThresholdParameter {
|
---|
75 | get { return (IFixedValueParameter<DoubleValue>)Parameters[NodeImpactThresholdParameterName]; }
|
---|
76 | }
|
---|
77 | public IFixedValueParameter<BoolValue> PruneOnlyZeroImpactNodesParameter {
|
---|
78 | get { return (IFixedValueParameter<BoolValue>)Parameters[PruneOnlyZeroImpactNodesParameterName]; }
|
---|
79 | }
|
---|
80 | public ILookupParameter<DoubleLimit> EstimationLimitsParameter {
|
---|
81 | get { return (ILookupParameter<DoubleLimit>)Parameters[EstimationLimitsParameterName]; }
|
---|
82 | }
|
---|
83 | public ILookupParameter<ISymbolicDataAnalysisExpressionTreeInterpreter> InterpreterParameter {
|
---|
84 | get { return (ILookupParameter<ISymbolicDataAnalysisExpressionTreeInterpreter>)Parameters[InterpreterParameterName]; }
|
---|
85 | }
|
---|
86 | #endregion
|
---|
87 |
|
---|
88 | #region properties
|
---|
89 | public ISymbolicDataAnalysisSolutionImpactValuesCalculator ImpactValuesCalculator {
|
---|
90 | get { return ImpactValuesCalculatorParameter.Value; }
|
---|
91 | set { ImpactValuesCalculatorParameter.Value = value; }
|
---|
92 | }
|
---|
93 | public bool PruneOnlyZeroImpactNodes {
|
---|
94 | get { return PruneOnlyZeroImpactNodesParameter.Value.Value; }
|
---|
95 | set { PruneOnlyZeroImpactNodesParameter.Value.Value = value; }
|
---|
96 | }
|
---|
97 | public double NodeImpactThreshold {
|
---|
98 | get { return NodeImpactThresholdParameter.Value.Value; }
|
---|
99 | set { NodeImpactThresholdParameter.Value.Value = value; }
|
---|
100 | }
|
---|
101 | #endregion
|
---|
102 |
|
---|
103 | [StorableConstructor]
|
---|
104 | protected SymbolicDataAnalysisExpressionPruningOperator(bool deserializing) : base(deserializing) { }
|
---|
105 | protected SymbolicDataAnalysisExpressionPruningOperator(SymbolicDataAnalysisExpressionPruningOperator original, Cloner cloner)
|
---|
106 | : base(original, cloner) { }
|
---|
107 |
|
---|
108 | protected SymbolicDataAnalysisExpressionPruningOperator(ISymbolicDataAnalysisSolutionImpactValuesCalculator impactValuesCalculator) {
|
---|
109 | #region add parameters
|
---|
110 | Parameters.Add(new LookupParameter<IDataAnalysisProblemData>(ProblemDataParameterName));
|
---|
111 | Parameters.Add(new LookupParameter<ISymbolicDataAnalysisModel>(SymbolicDataAnalysisModelParameterName));
|
---|
112 | Parameters.Add(new LookupParameter<IntRange>(FitnessCalculationPartitionParameterName));
|
---|
113 | Parameters.Add(new LookupParameter<IntValue>(PrunedSubtreesParameterName, "A counter of how many subtrees were replaced."));
|
---|
114 | Parameters.Add(new LookupParameter<IntValue>(PrunedTreesParameterName, "A counter of how many trees were pruned."));
|
---|
115 | Parameters.Add(new FixedValueParameter<BoolValue>(PruneOnlyZeroImpactNodesParameterName, "Specify whether or not only zero impact nodes should be pruned."));
|
---|
116 | Parameters.Add(new FixedValueParameter<DoubleValue>(NodeImpactThresholdParameterName, "Specifies an impact value threshold below which nodes should be pruned."));
|
---|
117 | Parameters.Add(new LookupParameter<DoubleLimit>(EstimationLimitsParameterName));
|
---|
118 | Parameters.Add(new LookupParameter<ISymbolicDataAnalysisExpressionTreeInterpreter>(InterpreterParameterName));
|
---|
119 | Parameters.Add(new LookupParameter<ISymbolicExpressionTree>(SymbolicExpressionTreeParameterName));
|
---|
120 | Parameters.Add(new LookupParameter<DoubleValue>(QualityParameterName));
|
---|
121 | Parameters.Add(new ValueParameter<ISymbolicDataAnalysisSolutionImpactValuesCalculator>(ImpactValuesCalculatorParameterName, impactValuesCalculator));
|
---|
122 | #endregion
|
---|
123 | }
|
---|
124 |
|
---|
125 | protected abstract ISymbolicDataAnalysisModel CreateModel(ISymbolicExpressionTree tree, ISymbolicDataAnalysisExpressionTreeInterpreter interpreter, IDataAnalysisProblemData problemData, DoubleLimit estimationLimits);
|
---|
126 |
|
---|
127 | protected abstract double Evaluate(IDataAnalysisModel model);
|
---|
128 |
|
---|
129 | public override IOperation Apply() {
|
---|
130 | var tree = SymbolicExpressionTreeParameter.ActualValue;
|
---|
131 | var problemData = ProblemDataParameter.ActualValue;
|
---|
132 | var fitnessCalculationPartition = FitnessCalculationPartitionParameter.ActualValue;
|
---|
133 | var estimationLimits = EstimationLimitsParameter.ActualValue;
|
---|
134 | var interpreter = InterpreterParameter.ActualValue;
|
---|
135 |
|
---|
136 | var model = CreateModel(tree, interpreter, problemData, estimationLimits);
|
---|
137 | var nodes = tree.Root.GetSubtree(0).GetSubtree(0).IterateNodesPrefix().ToList();
|
---|
138 | var rows = Enumerable.Range(fitnessCalculationPartition.Start, fitnessCalculationPartition.Size);
|
---|
139 | var prunedSubtrees = 0;
|
---|
140 | var prunedTrees = 0;
|
---|
141 |
|
---|
142 | double quality = Evaluate(model);
|
---|
143 |
|
---|
144 | for (int i = 0; i < nodes.Count; ++i) {
|
---|
145 | var node = nodes[i];
|
---|
146 | if (node is ConstantTreeNode) continue;
|
---|
147 |
|
---|
148 | double impactValue, replacementValue;
|
---|
149 | ImpactValuesCalculator.CalculateImpactAndReplacementValues(model, node, problemData, rows, out impactValue, out replacementValue, quality);
|
---|
150 |
|
---|
151 | if (PruneOnlyZeroImpactNodes && !impactValue.IsAlmost(0.0)) continue;
|
---|
152 | if (!PruneOnlyZeroImpactNodes && impactValue > NodeImpactThreshold) continue;
|
---|
153 |
|
---|
154 | var constantNode = (ConstantTreeNode)node.Grammar.GetSymbol("Constant").CreateTreeNode();
|
---|
155 | constantNode.Value = replacementValue;
|
---|
156 |
|
---|
157 | ReplaceWithConstant(node, constantNode);
|
---|
158 | i += node.GetLength() - 1; // skip subtrees under the node that was folded
|
---|
159 |
|
---|
160 | quality -= impactValue;
|
---|
161 | prunedSubtrees++;
|
---|
162 | }
|
---|
163 |
|
---|
164 | if (prunedSubtrees > 0) prunedTrees = 1;
|
---|
165 | PrunedSubtreesParameter.ActualValue = new IntValue(prunedSubtrees);
|
---|
166 | PrunedTreesParameter.ActualValue = new IntValue(prunedTrees);
|
---|
167 | QualityParameter.ActualValue.Value = quality;
|
---|
168 |
|
---|
169 | return base.Apply();
|
---|
170 | }
|
---|
171 |
|
---|
172 | protected static void ReplaceWithConstant(ISymbolicExpressionTreeNode original, ISymbolicExpressionTreeNode replacement) {
|
---|
173 | var parent = original.Parent;
|
---|
174 | var i = parent.IndexOfSubtree(original);
|
---|
175 | parent.RemoveSubtree(i);
|
---|
176 | parent.InsertSubtree(i, replacement);
|
---|
177 | }
|
---|
178 | }
|
---|
179 | }
|
---|