1 | #region License Information
|
---|
2 | /* HeuristicLab
|
---|
3 | * Copyright (C) 2002-2015 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 System.Collections.Generic;
|
---|
24 | using System.Linq;
|
---|
25 |
|
---|
26 | using HeuristicLab.Common;
|
---|
27 | using HeuristicLab.Core;
|
---|
28 | using HeuristicLab.Data;
|
---|
29 | using HeuristicLab.Parameters;
|
---|
30 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
31 | using HeuristicLab.Random;
|
---|
32 | using HeuristicLab.Selection; |
---|
33 | |
---|
34 | namespace HeuristicLab.Problems.ProgramSynthesis { |
---|
35 | /// <summary>
|
---|
36 | /// A lexicase selection operator which considers all successful evaluated training cases for selection.
|
---|
37 | /// </summary>
|
---|
38 | [Item("LexicaseSelector", "A lexicase selection operator which considers all successful evaluated training cases for selection.")]
|
---|
39 | [StorableClass]
|
---|
40 | public sealed class LexicaseSelector : StochasticSingleObjectiveSelector, ICaseSingleObjectiveSelector {
|
---|
41 | public ILookupParameter<ItemArray<DoubleArray>> CaseQualitiesParameter {
|
---|
42 | get { return (ILookupParameter<ItemArray<DoubleArray>>)Parameters[IntegerVectorPushProblem.CaseQualitiesScopeParameterName]; }
|
---|
43 | }
|
---|
44 |
|
---|
45 | [StorableConstructor]
|
---|
46 | private LexicaseSelector(bool deserializing) : base(deserializing) { }
|
---|
47 | private LexicaseSelector(LexicaseSelector original, Cloner cloner) : base(original, cloner) { }
|
---|
48 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
49 | return new LexicaseSelector(this, cloner);
|
---|
50 | }
|
---|
51 |
|
---|
52 | public LexicaseSelector() {
|
---|
53 | Parameters.Add(new ScopeTreeLookupParameter<DoubleArray>(
|
---|
54 | IntegerVectorPushProblem.CaseQualitiesScopeParameterName,
|
---|
55 | "The quality of every single training case for each individual."));
|
---|
56 | }
|
---|
57 |
|
---|
58 |
|
---|
59 | protected override IScope[] Select(List<IScope> scopes) {
|
---|
60 | var count = NumberOfSelectedSubScopesParameter.ActualValue.Value;
|
---|
61 | var copy = CopySelectedParameter.Value.Value;
|
---|
62 | var random = RandomParameter.ActualValue;
|
---|
63 | var maximization = MaximizationParameter.ActualValue.Value;
|
---|
64 | var caseQualities = CaseQualitiesParameter.ActualValue.ToList();
|
---|
65 |
|
---|
66 | var selected = Apply(
|
---|
67 | scopes,
|
---|
68 | count,
|
---|
69 | copy,
|
---|
70 | maximization,
|
---|
71 | random,
|
---|
72 | caseQualities);
|
---|
73 |
|
---|
74 | return selected;
|
---|
75 | }
|
---|
76 |
|
---|
77 | public static IScope[] Apply(
|
---|
78 | List<IScope> scopes,
|
---|
79 | int count,
|
---|
80 | bool copy,
|
---|
81 | bool maximization,
|
---|
82 | IRandom random,
|
---|
83 | List<DoubleArray> caseQualities) {
|
---|
84 |
|
---|
85 | for (var i = 0; i < caseQualities.Count; i++) {
|
---|
86 | if (caseQualities[i].Length == 0) {
|
---|
87 | scopes.RemoveAt(i);
|
---|
88 | caseQualities.RemoveAt(i);
|
---|
89 | }
|
---|
90 | }
|
---|
91 |
|
---|
92 | var qualitiesLength = caseQualities[0].Length;
|
---|
93 |
|
---|
94 | if (caseQualities.Any(x => x.Length != qualitiesLength)) {
|
---|
95 | throw new ArgumentException("Not all case qualities have the same length");
|
---|
96 | }
|
---|
97 |
|
---|
98 | var selected = new IScope[count];
|
---|
99 |
|
---|
100 | var candidates = new List<int>(caseQualities.Count);
|
---|
101 | for (var i = 0; i < caseQualities.Count; i++) candidates.Add(i);
|
---|
102 |
|
---|
103 | var orderSource = new List<int>(qualitiesLength);
|
---|
104 | for (var i = 0; i < qualitiesLength; i++) orderSource.Add(i);
|
---|
105 |
|
---|
106 | for (var i = 0; i < count; i++) {
|
---|
107 | var index = LexicaseSelect(
|
---|
108 | caseQualities,
|
---|
109 | candidates,
|
---|
110 | orderSource.Shuffle(random),
|
---|
111 | random,
|
---|
112 | maximization);
|
---|
113 |
|
---|
114 | if (copy) {
|
---|
115 | selected[i] = (IScope)scopes[index].Clone();
|
---|
116 | } else {
|
---|
117 | selected[i] = scopes[index];
|
---|
118 | scopes.RemoveAt(index);
|
---|
119 | caseQualities.RemoveAt(index);
|
---|
120 | }
|
---|
121 | }
|
---|
122 |
|
---|
123 | return selected;
|
---|
124 | }
|
---|
125 |
|
---|
126 | private static int LexicaseSelect(
|
---|
127 | List<DoubleArray> caseQualities,
|
---|
128 | List<int> candidates,
|
---|
129 | IEnumerable<int> order,
|
---|
130 | IRandom random,
|
---|
131 | bool maximization) {
|
---|
132 |
|
---|
133 | foreach (var curCase in order) {
|
---|
134 | var nextCandidates = new List<int>();
|
---|
135 |
|
---|
136 | var best = maximization
|
---|
137 | ? double.NegativeInfinity
|
---|
138 | : double.PositiveInfinity;
|
---|
139 |
|
---|
140 | for (var i = 0; i < candidates.Count; i++) {
|
---|
141 | var candidate = candidates[i];
|
---|
142 | var caseQuality = caseQualities[candidate][curCase];
|
---|
143 |
|
---|
144 | if (caseQuality.IsAlmost(best)) {
|
---|
145 | // if the individuals is as good as the best one, add it
|
---|
146 | nextCandidates.Add(candidate);
|
---|
147 | } else if (
|
---|
148 | (maximization && (caseQuality > best)) ||
|
---|
149 | (!maximization && (caseQuality < best))) {
|
---|
150 | // if the individual is better than the best one, remove all previous candidates and add the new one
|
---|
151 | nextCandidates.Clear();
|
---|
152 | nextCandidates.Add(candidate);
|
---|
153 | // also set the next best quality value
|
---|
154 | best = caseQuality;
|
---|
155 | }
|
---|
156 | // else {do nothing}
|
---|
157 | }
|
---|
158 |
|
---|
159 | if (nextCandidates.Count == 1) {
|
---|
160 | return nextCandidates[0];
|
---|
161 | }
|
---|
162 |
|
---|
163 | if (nextCandidates.Count < 1) {
|
---|
164 | return candidates.SampleRandom(random);
|
---|
165 | }
|
---|
166 |
|
---|
167 | candidates = nextCandidates;
|
---|
168 | }
|
---|
169 |
|
---|
170 | return candidates.Count == 1
|
---|
171 | ? candidates[0]
|
---|
172 | : candidates.SampleRandom(random);
|
---|
173 | }
|
---|
174 | }
|
---|
175 | }
|
---|