#region License Information /* HeuristicLab * Copyright (C) 2002-2008 Heuristic and Evolutionary Algorithms Laboratory (HEAL) * * This file is part of HeuristicLab. * * HeuristicLab is free software: you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation, either version 3 of the License, or * (at your option) any later version. * * HeuristicLab is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with HeuristicLab. If not, see . */ #endregion using System; using System.Collections.Generic; using System.Linq; using System.Text; using HeuristicLab.Core; using HeuristicLab.Operators; using HeuristicLab.Random; using HeuristicLab.Data; using HeuristicLab.Constraints; using System.Diagnostics; namespace HeuristicLab.GP { /// /// Implementation of a homologous uniform crossover operator as described in: /// R. Poli and W. B. Langdon. On the Search Properties of Different Crossover Operators in Genetic Programming. /// In Proceedings of Genetic Programming '98, Madison, Wisconsin, 1998. /// public class UniformCrossover : OperatorBase { public override string Description { get { return @"Uniform crossover as defined by Poli and Langdon"; } } public UniformCrossover() : base() { AddVariableInfo(new VariableInfo("Random", "Pseudo random number generator", typeof(MersenneTwister), VariableKind.In)); AddVariableInfo(new VariableInfo("OperatorLibrary", "The operator library containing all available operators", typeof(GPOperatorLibrary), VariableKind.In)); AddVariableInfo(new VariableInfo("FunctionTree", "The tree to mutate", typeof(IFunctionTree), VariableKind.In | VariableKind.New)); AddVariableInfo(new VariableInfo("TreeSize", "The size (number of nodes) of the tree", typeof(IntData), VariableKind.In | VariableKind.New)); AddVariableInfo(new VariableInfo("TreeHeight", "The height of the tree", typeof(IntData), VariableKind.In | VariableKind.New)); } public override IOperation Apply(IScope scope) { MersenneTwister random = GetVariableValue("Random", scope, true); GPOperatorLibrary opLibrary = GetVariableValue("OperatorLibrary", scope, true); TreeGardener gardener = new TreeGardener(random, opLibrary); if ((scope.SubScopes.Count % 2) != 0) throw new InvalidOperationException("Number of parents is not even"); CompositeOperation initOperations = new CompositeOperation(); int crossoverEvents = scope.SubScopes.Count / 2; for (int i = 0; i < crossoverEvents; i++) { IScope parent1 = scope.SubScopes[0]; scope.RemoveSubScope(parent1); IScope parent2 = scope.SubScopes[0]; scope.RemoveSubScope(parent2); IScope child0 = new Scope((i * 2).ToString()); IScope child1 = new Scope((i * 2 + 1).ToString()); Cross(scope, random, gardener, parent1, parent2, child0, child1); scope.AddSubScope(child0); scope.AddSubScope(child1); } return null; } private void Cross(IScope scope, MersenneTwister random, TreeGardener gardener, IScope parent1, IScope parent2, IScope child0, IScope child1) { IFunctionTree childTree0; IFunctionTree childTree1; Cross(random, gardener, parent1, parent2, out childTree0, out childTree1); Debug.Assert(gardener.IsValidTree(childTree0)); Debug.Assert(gardener.IsValidTree(childTree1)); child0.AddVariable(new HeuristicLab.Core.Variable(scope.TranslateName("FunctionTree"), childTree0)); child0.AddVariable(new HeuristicLab.Core.Variable(scope.TranslateName("TreeSize"), new IntData(childTree0.Size))); child0.AddVariable(new HeuristicLab.Core.Variable(scope.TranslateName("TreeHeight"), new IntData(childTree0.Height))); child1.AddVariable(new HeuristicLab.Core.Variable(scope.TranslateName("FunctionTree"), childTree1)); child1.AddVariable(new HeuristicLab.Core.Variable(scope.TranslateName("TreeSize"), new IntData(childTree1.Size))); child1.AddVariable(new HeuristicLab.Core.Variable(scope.TranslateName("TreeHeight"), new IntData(childTree1.Height))); } private void Cross(MersenneTwister random, TreeGardener gardener, IScope f, IScope g, out IFunctionTree child0, out IFunctionTree child1) { IFunctionTree tree0 = f.GetVariableValue("FunctionTree", false); int tree0Height = f.GetVariableValue("TreeHeight", false).Data; int tree0Size = f.GetVariableValue("TreeSize", false).Data; IFunctionTree tree1 = g.GetVariableValue("FunctionTree", false); int tree1Height = g.GetVariableValue("TreeHeight", false).Data; int tree1Size = g.GetVariableValue("TreeSize", false).Data; List allowedCrossOverPoints = GetCrossOverPoints(gardener, tree0, tree1); foreach (CrossoverPoint p in allowedCrossOverPoints) { Debug.Assert(gardener.GetAllowedSubFunctions(p.parent0.Function, p.childIndex).Contains(p.parent1.SubTrees[p.childIndex].Function)); Debug.Assert(gardener.GetAllowedSubFunctions(p.parent1.Function, p.childIndex).Contains(p.parent0.SubTrees[p.childIndex].Function)); } // iterate through the list of crossover points and swap nodes with p=0.5 foreach (CrossoverPoint crossoverPoint in allowedCrossOverPoints) { if (random.NextDouble() < 0.5) { IFunctionTree parent0 = crossoverPoint.parent0; IFunctionTree parent1 = crossoverPoint.parent1; IFunctionTree branch0 = crossoverPoint.parent0.SubTrees[crossoverPoint.childIndex]; IFunctionTree branch1 = crossoverPoint.parent1.SubTrees[crossoverPoint.childIndex]; // if we are at an internal node of the common region swap only the node but not the subtrees if (branch0.SubTrees.Count == branch1.SubTrees.Count) { if (parent0 != null) { Debug.Assert(parent1 != null); Debug.Assert(branch0 != null); Debug.Assert(branch0 != null); Debug.Assert(gardener.GetAllowedSubFunctions(parent0.Function, crossoverPoint.childIndex).Contains(branch1.Function)); Debug.Assert(gardener.GetAllowedSubFunctions(parent1.Function, crossoverPoint.childIndex).Contains(branch0.Function)); // we are not at the root => exchange the branches in the parent parent0.RemoveSubTree(crossoverPoint.childIndex); parent1.RemoveSubTree(crossoverPoint.childIndex); parent0.InsertSubTree(crossoverPoint.childIndex, branch1); parent1.InsertSubTree(crossoverPoint.childIndex, branch0); } // always exchange all children List branch0Children = new List(branch0.SubTrees); // create backup lists List branch1Children = new List(branch1.SubTrees); while (branch0.SubTrees.Count > 0) branch0.RemoveSubTree(0); // remove all children while (branch1.SubTrees.Count > 0) branch1.RemoveSubTree(0); foreach (IFunctionTree subTree in branch1Children) { Debug.Assert(gardener.GetAllowedSubFunctions(branch0.Function, branch0.SubTrees.Count).Contains(subTree.Function)); branch0.AddSubTree(subTree); // append children of branch1 to branch0 } foreach (IFunctionTree subTree in branch0Children) { Debug.Assert(gardener.GetAllowedSubFunctions(branch1.Function, branch1.SubTrees.Count).Contains(subTree.Function)); branch1.AddSubTree(subTree); // and vice versa } } else { // If we are at a node at the border of the common region then exchange the whole branch. // If we are at the root node and the number of children is already different we can't do anything now but // at the end either tree0 or tree1 must be returned with p=0.5. // However if we are not at the root => exchange the branches in the parent if (parent0 != null) { Debug.Assert(parent1 != null); Debug.Assert(branch0 != null); Debug.Assert(branch1 != null); Debug.Assert(gardener.GetAllowedSubFunctions(parent0.Function, crossoverPoint.childIndex).Contains(branch1.Function)); Debug.Assert(gardener.GetAllowedSubFunctions(parent1.Function, crossoverPoint.childIndex).Contains(branch0.Function)); parent0.RemoveSubTree(crossoverPoint.childIndex); parent1.RemoveSubTree(crossoverPoint.childIndex); parent0.InsertSubTree(crossoverPoint.childIndex, branch1); parent1.InsertSubTree(crossoverPoint.childIndex, branch0); } } } } child0 = tree0; child1 = tree1; } class CrossoverPoint { public IFunctionTree parent0; public IFunctionTree parent1; public int childIndex; } private List GetCrossOverPoints(TreeGardener gardener, IFunctionTree branch0, IFunctionTree branch1) { List results = new List(); if (branch0.SubTrees.Count != branch1.SubTrees.Count) return results; for (int i = 0; i < branch0.SubTrees.Count; i++) { // if the branches fit to the parent if (gardener.GetAllowedSubFunctions(branch0.Function, i).Contains(branch1.SubTrees[i].Function) && gardener.GetAllowedSubFunctions(branch1.Function, i).Contains(branch0.SubTrees[i].Function)) { // if the point is at the border of the common region we don't care about the children // however if the point is not on the border of the common region we also have to check if // the children of the branches fit together bool fit = true; if (branch0.SubTrees[i].SubTrees.Count == branch1.SubTrees[i].SubTrees.Count) { for (int j = 0; j < branch0.SubTrees[i].SubTrees.Count; j++) { fit = fit & gardener.GetAllowedSubFunctions(branch0.SubTrees[i].Function, j).Contains(branch1.SubTrees[i].SubTrees[j].Function); fit = fit & gardener.GetAllowedSubFunctions(branch1.SubTrees[i].Function, j).Contains(branch0.SubTrees[i].SubTrees[j].Function); } } if (fit) { CrossoverPoint p = new CrossoverPoint(); p.childIndex = i; p.parent0 = branch0; p.parent1 = branch1; results.Add(p); } } results.AddRange(GetCrossOverPoints(gardener, branch0.SubTrees[i], branch1.SubTrees[i])); } return results; } } }