Free cookie consent management tool by TermsFeed Policy Generator

Changeset 11895 for branches


Ignore:
Timestamp:
02/05/15 07:03:15 (10 years ago)
Author:
gkronber
Message:

#2283: constant opt, expressioncompiler, autodiff, fixes in GP solvers

Location:
branches/HeuristicLab.Problems.GrammaticalOptimization
Files:
2 added
10 edited

Legend:

Unmodified
Added
Removed

  • branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Algorithms.GeneticProgramming/GenericSymbExprEvaluator.cs

    r11857 r11895  
    66using HeuristicLab.Operators;
    77using HeuristicLab.Parameters;
     8using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
    89using HeuristicLab.PluginInfrastructure;
    910
    1011namespace HeuristicLab.Problems.GrammaticalOptimization {
    11   // this type is not storable because we use a func<ITree,double> for evaluation, which references back to the original grammatical optimization problem
    1212  [NonDiscoverableType]
     13  [StorableClass]
    1314  [Item("GenericSymbExprEvaluator", "Evaluator for grammatical optimization problems (using a symbolic expression tree encoding).")]
    1415  public class GenericSymbExprEvaluator : SingleSuccessorOperator, IGenericSymbExprEvaluator {
    … …  
    2223    }
    2324
     25    // cannot save these (eval won't work when loaded from file
    2426    private Func<string, double> evalFunc;
    2527    private Func<ISymbolicExpressionTree, string> toStringFunc;
    2628
     29    [StorableConstructor]
     30    private GenericSymbExprEvaluator(bool deserializing) : base(deserializing) { }
    2731    public GenericSymbExprEvaluator(GenericSymbExprEvaluator original, Cloner cloner)
    2832      : base(original, cloner) {

  • branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Algorithms.GeneticProgramming/HeuristicLab.Algorithms.GeneticProgramming.csproj

    r11857 r11895  
    3636    <Reference Include="HeuristicLab.Algorithms.OffspringSelectionGeneticAlgorithm-3.3">
    3737      <HintPath>..\..\..\trunk\sources\bin\HeuristicLab.Algorithms.OffspringSelectionGeneticAlgorithm-3.3.dll</HintPath>
     38    </Reference>
     39    <Reference Include="HeuristicLab.Analysis-3.3, Version=3.3.0.0, Culture=neutral, PublicKeyToken=ba48961d6f65dcec, processorArchitecture=MSIL">
     40      <SpecificVersion>False</SpecificVersion>
     41      <HintPath>..\..\..\trunk\sources\bin\HeuristicLab.Analysis-3.3.dll</HintPath>
    3842    </Reference>
    3943    <Reference Include="HeuristicLab.Collections-3.3">
    … …  
    7579      <HintPath>..\..\..\trunk\sources\bin\HeuristicLab.Selection-3.3.dll</HintPath>
    7680    </Reference>
     81    <Reference Include="HeuristicLab.SequentialEngine-3.3">
     82      <HintPath>..\..\..\..\..\Program Files\HeuristicLab 3.3\HeuristicLab.SequentialEngine-3.3.dll</HintPath>
     83    </Reference>
    7784    <Reference Include="System" />
    7885    <Reference Include="System.Core" />
    … …  
    8087  </ItemGroup>
    8188  <ItemGroup>
     89    <Compile Include="IGPSolver.cs" />
    8290    <Compile Include="GenericSymbExprEvaluator.cs" />
    8391    <Compile Include="GenericSymbExprGrammar.cs" />

  • branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Algorithms.GeneticProgramming/OffspringSelectionGP.cs

    r11851 r11895  
    11using System;
     2using System.IO;
     3using System.IO.Compression;
    24using System.Linq;
    35using System.Threading;
    46using HeuristicLab.Algorithms.GrammaticalOptimization;
     7using HeuristicLab.Analysis;
     8using HeuristicLab.Common;
    59using HeuristicLab.Data;
    610using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
    7 using HeuristicLab.Optimization;
     11using HeuristicLab.Persistence.Default.Xml;
    812using HeuristicLab.Problems.GrammaticalOptimization;
    9 using HeuristicLab.Selection;
    10 using HeuristicLab.Algorithms.GeneticAlgorithm;
    1113
    1214namespace HeuristicLab.Algorithms.GeneticProgramming {
    13   public class OffspringSelectionGP : SolverBase {
     15  public class OffspringSelectionGP : SolverBase, IGPSolver {
    1416    public int PopulationSize { get; set; }
    1517    public double MutationRate { get; set; }
    … …  
    1921    private readonly ISymbolicExpressionTreeProblem problem;
    2022    private readonly Random random;
     23    private readonly bool saveAlg;
    2124
    22     public OffspringSelectionGP(ISymbolicExpressionTreeProblem problem, Random random) {
     25    public OffspringSelectionGP(ISymbolicExpressionTreeProblem problem, Random random, bool saveAlg = false) {
    2326      this.problem = problem;
    2427      this.random = random;
    … …  
    2831      MaxSolutionSize = 100;
    2932      MaxSolutionDepth = 17;
     33      this.saveAlg = saveAlg;
    3034    }
    3135
    … …  
    3337      var hlProblem = new GenericSymbExprProblem(problem);
    3438      var onEvalLocker = new object();
    35       hlProblem.Evaluator.SolutionEvaluated += (sentence, quality) => {
    36         // raise solution evaluated event for each GP solution, don't scale quality to 0..1
    37         // need to synchronize in case we are using a parallel engine
    38         lock (onEvalLocker) {
    39           OnSolutionEvaluated(sentence, quality);
    40         }
    41       };
    4239      hlProblem.MaximumSymbolicExpressionTreeLength.Value = MaxSolutionSize;
    4340      hlProblem.MaximumSymbolicExpressionTreeDepth.Value = MaxSolutionDepth;
    … …  
    4542      using (var wh = new AutoResetEvent(false)) {
    4643        var osga = new OffspringSelectionGeneticAlgorithm.OffspringSelectionGeneticAlgorithm();
    47         osga.Engine = new ParallelEngine.ParallelEngine();
     44        // osga.Engine = new ParallelEngine.ParallelEngine();
     45        osga.Engine = new SequentialEngine.SequentialEngine();
    4846        osga.ExceptionOccurred += (sender, args) => { Console.WriteLine(args.Value.Message); wh.Set(); };
    4947        osga.Stopped += (sender, args) => { wh.Set(); };
     48
     49        int numEvals = 0;
     50        hlProblem.Evaluator.SolutionEvaluated += (sentence, quality) => {
     51          // raise solution evaluated event for each GP solution, don't scale quality to 0..1
     52          // need to synchronize in case we are using a parallel engine
     53          lock (onEvalLocker) {
     54            OnSolutionEvaluated(sentence, quality);
     55
     56            // stop when maxEvals has been reached
     57            if (numEvals++ >= maxEvaluations) {
     58              osga.Stop();
     59            }
     60          }
     61        };
     62
    5063
    5164        osga.Problem = hlProblem;
    … …  
    5972        osga.Crossover = osga.CrossoverParameter.ValidValues.Single(op => op.Name == "SubtreeSwappingCrossover");
    6073        osga.Selector = osga.SelectorParameter.ValidValues.Single(op => op.Name == "GenderSpecificSelection");
     74        var multiAnalzer = (MultiAnalyzer)osga.Analyzer;
     75        multiAnalzer.Operators.Add(new BestSymbolicExpressionTreeAnalyzer());
    6176
    6277        osga.PopulationSize.Value = PopulationSize;
    63         osga.MaximumGenerations.Value = 100000; // some very large value (we stop based on evaluations)
    64         osga.MaximumSelectionPressure.Value = 1000;
     78        osga.MaximumGenerations.Value = 1000000; // some very large value (we stop based on evaluations)
     79        osga.MaximumSelectionPressure.Value = 1000000;
    6580        osga.MaximumEvaluatedSolutions.Value = maxEvaluations;
    6681        osga.MutationProbability.Value = MutationRate;
     82        osga.ComparisonFactorLowerBound.Value = 1.0;
     83        osga.ComparisonFactorUpperBound.Value = 1.0;
     84        osga.SuccessRatio.Value = 1.0;
    6785
    6886        osga.SetSeedRandomly = new BoolValue(false);
    … …  
    7391
    7492        wh.WaitOne();
     93
     94
     95        if (saveAlg) {
     96          var path = @"C:\Users\P24581\Desktop";
     97          var fileName = string.Format("osgp-{0}{1:D2}{2:D2}{3:D2}{4:D2}.hl", DateTime.Now.Year, DateTime.Now.Month, DateTime.Now.Day, DateTime.Now.Hour, DateTime.Now.Minute);
     98          var fullPath = Path.Combine(path, fileName);
     99          HeuristicLab.Persistence.Core.ConfigurationService.Instance.LoadSettings();
     100          XmlGenerator.Serialize(osga, fullPath, CompressionLevel.Fastest);
     101        }
    75102      }
    76103    }

  • branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Algorithms.GeneticProgramming/StandardGP.cs

    r11851 r11895  
    11using System;
     2using System.IO;
     3using System.IO.Compression;
    24using System.Linq;
    35using System.Threading;
    46using HeuristicLab.Algorithms.GrammaticalOptimization;
     7using HeuristicLab.Common;
    58using HeuristicLab.Data;
    69using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
    7 using HeuristicLab.Optimization;
     10using HeuristicLab.Persistence.Default.Xml;
    811using HeuristicLab.Problems.GrammaticalOptimization;
    912using HeuristicLab.Selection;
    10 using HeuristicLab.Algorithms.GeneticAlgorithm;
    1113
    1214namespace HeuristicLab.Algorithms.GeneticProgramming {
    13   public class StandardGP : SolverBase {
     15  public class StandardGP : SolverBase, IGPSolver {
    1416    public int PopulationSize { get; set; }
    1517    public double MutationRate { get; set; }
    … …  
    2022    private readonly ISymbolicExpressionTreeProblem problem;
    2123    private readonly Random random;
     24    private readonly bool saveAlg;
    2225
    23     public StandardGP(ISymbolicExpressionTreeProblem problem, Random random) {
     26    public StandardGP(ISymbolicExpressionTreeProblem problem, Random random, bool saveAlg = false) {
    2427      this.problem = problem;
    2528      this.random = random;
    … …  
    3033      MaxSolutionSize = 100;
    3134      MaxSolutionDepth = 17;
     35      this.saveAlg = saveAlg;
    3236    }
    3337
    … …  
    3539      var hlProblem = new GenericSymbExprProblem(problem);
    3640      var onEvalLocker = new object();
    37       hlProblem.Evaluator.SolutionEvaluated += (sentence, quality) => {
    38         // raise solution evaluated event for each GP solution, don't scale quality to 0..1
    39         // need to synchronize in case we are using a parallel engine
    40         lock (onEvalLocker) {
    41           OnSolutionEvaluated(sentence, quality);
    42         }
    43       };
    4441      hlProblem.MaximumSymbolicExpressionTreeLength.Value = MaxSolutionSize;
    4542      hlProblem.MaximumSymbolicExpressionTreeDepth.Value = MaxSolutionDepth;
    … …  
    5148        ga.ExceptionOccurred += (sender, args) => { Console.WriteLine(args.Value.Message); wh.Set(); };
    5249        ga.Stopped += (sender, args) => { wh.Set(); };
     50
     51        int numEvals = 0;
     52        hlProblem.Evaluator.SolutionEvaluated += (sentence, quality) => {
     53          // raise solution evaluated event for each GP solution, don't scale quality to 0..1
     54          // need to synchronize in case we are using a parallel engine
     55          lock (onEvalLocker) {
     56            OnSolutionEvaluated(sentence, quality);
     57
     58            // stop when maxEvals has been reached
     59            if (numEvals++ >= maxEvaluations) {
     60              ga.Stop();
     61            }
     62          }
     63        };
     64
    5365
    5466        ga.Problem = hlProblem;
    … …  
    6678
    6779        ga.PopulationSize.Value = PopulationSize;
    68         ga.MaximumGenerations.Value = maxEvaluations / PopulationSize + 1; // one extra generation in case maxEvaluations is not a multiple of PopulationSize
     80        ga.MaximumGenerations.Value = 1000000; // very large value (we stop in the evaluate handler)
    6981        ga.MutationProbability.Value = MutationRate;
    7082
    … …  
    7688
    7789        wh.WaitOne();
     90
     91
     92        if (saveAlg) {
     93          var path = @"C:\Users\P24581\Desktop";
     94          var fileName = string.Format("osgp-{0}{1:D2}{2:D2}{3:D2}{4:D2}.hl", DateTime.Now.Year, DateTime.Now.Month, DateTime.Now.Day, DateTime.Now.Hour, DateTime.Now.Minute);
     95          var fullPath = Path.Combine(path, fileName);
     96          HeuristicLab.Persistence.Core.ConfigurationService.Instance.LoadSettings();
     97          XmlGenerator.Serialize(ga, fullPath, CompressionLevel.Fastest);
     98        }
    7899      }
    79100    }

  • branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Problems.GrammaticalOptimization.SymbReg/HeuristicLab.Problems.GrammaticalOptimization.SymbReg.csproj

    r11851 r11895  
    3131  </PropertyGroup>
    3232  <ItemGroup>
     33    <Reference Include="ALGLIB-3.7.0, Version=3.7.0.0, Culture=neutral, PublicKeyToken=ba48961d6f65dcec, processorArchitecture=MSIL">
     34      <SpecificVersion>False</SpecificVersion>
     35      <HintPath>..\..\..\trunk\sources\bin\ALGLIB-3.7.0.dll</HintPath>
     36    </Reference>
     37    <Reference Include="AutoDiff-1.0">
     38      <HintPath>..\..\..\trunk\sources\bin\AutoDiff-1.0.dll</HintPath>
     39    </Reference>
    3340    <Reference Include="HeuristicLab.Common-3.3">
    3441      <HintPath>..\..\..\trunk\sources\bin\HeuristicLab.Common-3.3.dll</HintPath>
    … …  
    3946    <Reference Include="HeuristicLab.Data-3.3">
    4047      <HintPath>..\..\..\trunk\sources\bin\HeuristicLab.Data-3.3.dll</HintPath>
     48    </Reference>
     49    <Reference Include="HeuristicLab.Encodings.SymbolicExpressionTreeEncoding-3.4, Version=3.4.0.0, Culture=neutral, PublicKeyToken=ba48961d6f65dcec, processorArchitecture=MSIL">
     50      <SpecificVersion>False</SpecificVersion>
     51      <HintPath>..\..\..\trunk\sources\bin\HeuristicLab.Encodings.SymbolicExpressionTreeEncoding-3.4.dll</HintPath>
    4152    </Reference>
    4253    <Reference Include="HeuristicLab.Problems.DataAnalysis-3.4">
    … …  
    5364  </ItemGroup>
    5465  <ItemGroup>
     66    <Compile Include="ExpressionCompiler.cs" />
    5567    <Compile Include="Properties\AssemblyInfo.cs" />
    5668    <Compile Include="SymbolicRegressionProblem.cs" />

  • branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Problems.GrammaticalOptimization.SymbReg/SymbolicRegressionProblem.cs

    r11851 r11895  
    11using System;
     2using System.CodeDom.Compiler;
    23using System.Collections.Generic;
     4using System.Diagnostics;
    35using System.Linq;
    46using System.Security;
    57using System.Security.AccessControl;
     8using System.Security.Authentication.ExtendedProtection.Configuration;
    69using System.Text;
     10using AutoDiff;
    711using HeuristicLab.Common;
     12using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
    813using HeuristicLab.Problems.DataAnalysis;
    914using HeuristicLab.Problems.Instances;
    … …  
    1217namespace HeuristicLab.Problems.GrammaticalOptimization.SymbReg {
    1318  // provides bridge to HL regression problem instances
    14   public class SymbolicRegressionProblem : IProblem {
     19  public class SymbolicRegressionProblem : ISymbolicExpressionTreeProblem {
    1520
    1621    private const string grammarString = @"
    1722        G(E):
    18         E -> V | V+E | V-E | V*E | V/E | (E) | C | C+E | C-E | C*E | C/E
     23        E -> V | C | V+E | V-E | V*E | V%E | (E) | C+E | C-E | C*E | C%E
    1924        C -> 0..9
    2025        V -> <variables>
    … …  
    2227    // C represents Koza-style ERC (the symbol is the index of the ERC), the values are initialized below
    2328
     29    // S .. Sum (+), N .. Neg. sum (-), P .. Product (*), D .. Division (%)
     30    private const string treeGrammarString = @"
     31        G(E):
     32        E -> V | C | S | N | P | D
     33        S -> EE | EEE
     34        N -> EE | EEE
     35        P -> EE | EEE
     36        D -> EE
     37        C -> 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9
     38        V -> <variables>
     39        ";
     40
     41    // when we use constants optimization we can completely ignore all constants by a simple strategy:
     42    // introduce a constant factor for each complete term
     43    // introduce a constant offset for each complete expression (including expressions in brackets)
     44    // e.g. 1*(2*a + b - 3 + 4) is the same as c0*a + c1*b + c2
     45
    2446    private readonly IGrammar grammar;
    2547
    2648    private readonly int N;
    27     private readonly double[][] x;
     49    private readonly double[,] x;
    2850    private readonly double[] y;
    2951    private readonly int d;
    … …  
    4062      this.d = problemData.AllowedInputVariables.Count();
    4163      if (d > 26) throw new NotSupportedException(); // we only allow single-character terminal symbols so far
    42       this.x = new double[N][];
     64      this.x = new double[N, d];
    4365      this.y = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, problemData.TrainingIndices).ToArray();
    4466
    4567      int i = 0;
    4668      foreach (var r in problemData.TrainingIndices) {
    47         x[i] = new double[d];
    4869        int j = 0;
    4970        foreach (var inputVariable in problemData.AllowedInputVariables) {
    50           x[i][j++] = problemData.Dataset.GetDoubleValue(inputVariable, r);
     71          x[i, j++] = problemData.Dataset.GetDoubleValue(inputVariable, r);
    5172        }
    5273        i++;
    … …  
    5879      char lastVar = Convert.ToChar(Convert.ToByte('a') + d - 1);
    5980      this.grammar = new Grammar(grammarString.Replace("<variables>", firstVar + " .. " + lastVar));
     81      this.TreeBasedGPGrammar = new Grammar(treeGrammarString.Replace("<variables>", firstVar + " .. " + lastVar));
    6082    }
    6183
    … …  
    7193
    7294    public double Evaluate(string sentence) {
     95      return OptimizeConstantsAndEvaluate(sentence);
     96    }
     97
     98    public double SimpleEvaluate(string sentence) {
    7399      var interpreter = new ExpressionInterpreter();
    74       return HeuristicLab.Common.Extensions.RSq(y, Enumerable.Range(0, N).Select(i => interpreter.Interpret(sentence, x[i], erc)));
     100      var rowData = new double[d];
     101      return HeuristicLab.Common.Extensions.RSq(y, Enumerable.Range(0, N).Select(i => {
     102        for (int j = 0; j < d; j++) rowData[j] = x[i, j];
     103        return interpreter.Interpret(sentence, rowData, erc);
     104      }));
    75105    }
    76106
    … …  
    83113    }
    84114
    85     public IEnumerable<Feature> GetFeatures(string phrase)
    86     {
     115    public IEnumerable<Feature> GetFeatures(string phrase) {
    87116      throw new NotImplementedException();
    88117    }
    89118
    90119
    91     /*
    92     public static double OptimizeConstants(string sentence) {
    93 
    94       List<AutoDiff.Variable> variables = new List<AutoDiff.Variable>();
    95       List<AutoDiff.Variable> parameters = new List<AutoDiff.Variable>();
    96       List<string> variableNames = new List<string>();
    97 
     120
     121    public double OptimizeConstantsAndEvaluate(string sentence) {
    98122      AutoDiff.Term func;
    99       if (!TryTransformToAutoDiff(sentence, 0, variables, parameters, out func))
    100         throw new NotSupportedException("Could not optimize constants of symbolic expression tree due to not supported symbols used in the tree.");
    101       if (variableNames.Count == 0) return 0.0;
    102 
    103       AutoDiff.IParametricCompiledTerm compiledFunc = AutoDiff.TermUtils.Compile(func, variables.ToArray(), parameters.ToArray());
    104 
    105       List<SymbolicExpressionTreeTerminalNode> terminalNodes = tree.Root.IterateNodesPrefix().OfType<SymbolicExpressionTreeTerminalNode>().ToList();
    106       double[] c = new double[variables.Count];
    107 
    108       {
    109         c[0] = 0.0;
    110         c[1] = 1.0;
    111         //extract inital constants
    112         int i = 2;
    113         foreach (var node in terminalNodes) {
    114           ConstantTreeNode constantTreeNode = node as ConstantTreeNode;
    115           VariableTreeNode variableTreeNode = node as VariableTreeNode;
    116           if (constantTreeNode != null)
    117             c[i++] = constantTreeNode.Value;
    118           else if (variableTreeNode != null)
    119             c[i++] = variableTreeNode.Weight;
    120         }
    121       }
    122       double[] originalConstants = (double[])c.Clone();
    123       double originalQuality = SymbolicRegressionSingleObjectivePearsonRSquaredEvaluator.Calculate(interpreter, tree, lowerEstimationLimit, upperEstimationLimit, problemData, rows, applyLinearScaling);
     123      int pos = 0;
     124
     125      var compiler = new ExpressionCompiler();
     126      Variable[] variables;
     127      Variable[] constants;
     128      compiler.Compile(sentence, out func, out variables, out constants);
     129
     130      // constants are manipulated
     131
     132      if (!constants.Any()) return SimpleEvaluate(sentence);
     133
     134      AutoDiff.IParametricCompiledTerm compiledFunc = func.Compile(constants, variables); // variate constants leave variables fixed to data
     135
     136      double[] c = constants.Select(_ => 1.0).ToArray(); // start with ones
    124137
    125138      alglib.lsfitstate state;
    … …  
    127140      int info;
    128141
    129       Dataset ds = problemData.Dataset;
    130       double[,] x = new double[rows.Count(), variableNames.Count];
    131       int row = 0;
    132       foreach (var r in rows) {
    133         for (int col = 0; col < variableNames.Count; col++) {
    134           x[row, col] = ds.GetDoubleValue(variableNames[col], r);
    135         }
    136         row++;
    137       }
    138       double[] y = ds.GetDoubleValues(problemData.TargetVariable, rows).ToArray();
    139142      int n = x.GetLength(0);
    140143      int m = x.GetLength(1);
    … …  
    144147      alglib.ndimensional_pgrad function_cx_1_grad = CreatePGrad(compiledFunc);
    145148
     149      const int maxIterations = 10;
    146150      try {
    147151        alglib.lsfitcreatefg(x, y, c, n, m, k, false, out state);
    … …  
    151155        alglib.lsfitresults(state, out info, out c, out rep);
    152156      } catch (ArithmeticException) {
    153         return originalQuality;
     157        return 0.0;
    154158      } catch (alglib.alglibexception) {
    155         return originalQuality;
     159        return 0.0;
    156160      }
    157161
    158162      //info == -7  => constant optimization failed due to wrong gradient
    159       if (info != -7) throw new ArgumentException();
     163      if (info == -7) throw new ArgumentException();
     164      {
     165        var rowData = new double[d];
     166        return HeuristicLab.Common.Extensions.RSq(y, Enumerable.Range(0, N).Select(i => {
     167          for (int j = 0; j < d; j++) rowData[j] = x[i, j];
     168          return compiledFunc.Evaluate(c, rowData);
     169        }));
     170      }
    160171    }
    161172
    … …  
    175186    }
    176187
    177     private static bool TryTransformToAutoDiff(string phrase, int symbolPos, List<AutoDiff.Variable> variables, List<AutoDiff.Variable> parameters, out AutoDiff.Term term)
    178     {
    179       var curSy = phrase[0];
    180       if () {
    181         var var = new AutoDiff.Variable();
    182         variables.Add(var);
    183         term = var;
    184         return true;
    185       }
    186       if (node.Symbol is Variable) {
    187         var varNode = node as VariableTreeNode;
    188         var par = new AutoDiff.Variable();
    189         parameters.Add(par);
    190         variableNames.Add(varNode.VariableName);
    191         var w = new AutoDiff.Variable();
    192         variables.Add(w);
    193         term = AutoDiff.TermBuilder.Product(w, par);
    194         return true;
    195       }
    196       if (node.Symbol is Addition) {
    197         List<AutoDiff.Term> terms = new List<Term>();
    198         foreach (var subTree in node.Subtrees) {
    199           AutoDiff.Term t;
    200           if (!TryTransformToAutoDiff(subTree, variables, parameters, variableNames, out t)) {
    201             term = null;
    202             return false;
    203           }
    204           terms.Add(t);
     188    public IGrammar TreeBasedGPGrammar { get; private set; }
     189    public string ConvertTreeToSentence(ISymbolicExpressionTree tree) {
     190      var sb = new StringBuilder();
     191
     192      TreeToSentence(tree.Root.GetSubtree(0).GetSubtree(0), sb);
     193
     194      return sb.ToString();
     195    }
     196
     197    private void TreeToSentence(ISymbolicExpressionTreeNode treeNode, StringBuilder sb) {
     198      if (treeNode.SubtreeCount == 0) {
     199        // terminal
     200        sb.Append(treeNode.Symbol.Name);
     201      } else {
     202        string op = string.Empty;
     203        switch (treeNode.Symbol.Name) {
     204          case "S": op = "+"; break;
     205          case "N": op = "-"; break;
     206          case "P": op = "*"; break;
     207          case "D": op = "%"; break;
     208          default: {
     209              Debug.Assert(treeNode.SubtreeCount == 1);
     210              break;
     211            }
    205212        }
    206         term = AutoDiff.TermBuilder.Sum(terms);
    207         return true;
    208       }
    209       if (node.Symbol is Subtraction) {
    210         List<AutoDiff.Term> terms = new List<Term>();
    211         for (int i = 0; i < node.SubtreeCount; i++) {
    212           AutoDiff.Term t;
    213           if (!TryTransformToAutoDiff(node.GetSubtree(i), variables, parameters, variableNames, out t)) {
    214             term = null;
    215             return false;
    216           }
    217           if (i > 0) t = -t;
    218           terms.Add(t);
     213        // nonterminal
     214        if (op == "+" || op == "-") sb.Append("(");
     215        TreeToSentence(treeNode.Subtrees.First(), sb);
     216        foreach (var subTree in treeNode.Subtrees.Skip(1)) {
     217          sb.Append(op);
     218          TreeToSentence(subTree, sb);
    219219        }
    220         term = AutoDiff.TermBuilder.Sum(terms);
    221         return true;
    222       }
    223       if (node.Symbol is Multiplication) {
    224         AutoDiff.Term a, b;
    225         if (!TryTransformToAutoDiff(node.GetSubtree(0), variables, parameters, variableNames, out a) ||
    226           !TryTransformToAutoDiff(node.GetSubtree(1), variables, parameters, variableNames, out b)) {
    227           term = null;
    228           return false;
    229         } else {
    230           List<AutoDiff.Term> factors = new List<Term>();
    231           foreach (var subTree in node.Subtrees.Skip(2)) {
    232             AutoDiff.Term f;
    233             if (!TryTransformToAutoDiff(subTree, variables, parameters, variableNames, out f)) {
    234               term = null;
    235               return false;
    236             }
    237             factors.Add(f);
    238           }
    239           term = AutoDiff.TermBuilder.Product(a, b, factors.ToArray());
    240           return true;
    241         }
    242       }
    243       if (node.Symbol is Division) {
    244         // only works for at least two subtrees
    245         AutoDiff.Term a, b;
    246         if (!TryTransformToAutoDiff(node.GetSubtree(0), variables, parameters, variableNames, out a) ||
    247           !TryTransformToAutoDiff(node.GetSubtree(1), variables, parameters, variableNames, out b)) {
    248           term = null;
    249           return false;
    250         } else {
    251           List<AutoDiff.Term> factors = new List<Term>();
    252           foreach (var subTree in node.Subtrees.Skip(2)) {
    253             AutoDiff.Term f;
    254             if (!TryTransformToAutoDiff(subTree, variables, parameters, variableNames, out f)) {
    255               term = null;
    256               return false;
    257             }
    258             factors.Add(1.0 / f);
    259           }
    260           term = AutoDiff.TermBuilder.Product(a, 1.0 / b, factors.ToArray());
    261           return true;
    262         }
    263       }
    264      
    265       if (node.Symbol is StartSymbol) {
    266         var alpha = new AutoDiff.Variable();
    267         var beta = new AutoDiff.Variable();
    268         variables.Add(beta);
    269         variables.Add(alpha);
    270         AutoDiff.Term branchTerm;
    271         if (TryTransformToAutoDiff(node.GetSubtree(0), variables, parameters, variableNames, out branchTerm)) {
    272           term = branchTerm * alpha + beta;
    273           return true;
    274         } else {
    275           term = null;
    276           return false;
    277         }
    278       }
    279       term = null;
    280       return false;
    281     }
    282      */
     220        if (op == "+" || op == "-") sb.Append(")");
     221      }
     222    }
    283223  }
    284224}

  • branches/HeuristicLab.Problems.GrammaticalOptimization/HeuristicLab.Problems.GrammaticalOptimization/ExpressionInterpreter.cs

    r11848 r11895  
    1717    // interprets sentences from L(G(Expr)):
    1818    // Expr -> Term { ('+' | '-' | '^' ) Term }
    19     // Term -> Fact { ('*' | '/') Fact }
     19    // Term -> Fact { ('*' | '%') Fact }
    2020    // Fact -> '!' Expr | '(' Expr ')' | Var | const
    2121    // Var -> 'a'..'z'
    2222    // const -> '0' .. '9'
    2323
     24    // uses protected division symbol %
    2425    // constants are Koza-style ephemeral random constants (ERC). for now we only allow up to 10 constants.
    2526    // The constant symbols '0' .. '9' are treated as ERC indices
    … …  
    3536      var d = new double[vars.Length];
    3637      for (int i = 0; i < vars.Length; i++) d[i] = vars[i] ? 1.0 : 0.0;
    37       return ! DoubleExtensions.IsAlmost(Expr(d, erc), 0);
     38      return !DoubleExtensions.IsAlmost(Expr(d, erc), 0);
    3839    }
    3940
    … …  
    7374        if (curSy == '+') {
    7475          NewSy();
    75           r += Expr(d, erc);
     76          r += Term(d, erc);
    7677        } else if (curSy == '-') {
    7778          NewSy();
    78           r -= Expr(d, erc);
     79          r -= Term(d, erc);
    7980        } else {
    8081          NewSy();
    81           var e = Expr(d, erc);
    82           r = Not(r) * e + r * Not(e); // xor = (!x AND y) OR (x AND !y)
     82          throw new NotImplementedException();
     83          // var e = Expr(d, erc);
     84          // r = Not(r) * e + r * Not(e); // xor = (!x AND y) OR (x AND !y)
    8385        }
    8486        curSy = CurSy();
    … …  
    9193      r = Fact(d, erc);
    9294      var curSy = CurSy();
    93       while (curSy == '*' || curSy == '/') {
     95      while (curSy == '*' || curSy == '%') {
    9496        if (curSy == '*') {
    9597          NewSy();
    96           r *= Term(d, erc);
     98          r *= Fact(d, erc);
    9799        } else {
    98100          NewSy();
    99           r /= Term(d, erc);
     101          var nom = Fact(d, erc);
     102          if (HeuristicLab.Common.Extensions.IsAlmost(nom, 0.0)) nom = 1.0;
     103          r /= nom;
    100104        }
    101105        curSy = CurSy();

  • branches/HeuristicLab.Problems.GrammaticalOptimization/Main/Program.cs

    r11865 r11895  
    33using System.Diagnostics;
    44using System.Globalization;
     5using System.Runtime.Remoting.Messaging;
    56using System.Text;
    67using System.Threading;
    … …  
    1314using HeuristicLab.Algorithms.GrammaticalOptimization;
    1415using HeuristicLab.Problems.GrammaticalOptimization;
     16using HeuristicLab.Problems.GrammaticalOptimization.SymbReg;
    1517using BoltzmannExplorationPolicy = HeuristicLab.Algorithms.Bandits.BanditPolicies.BoltzmannExplorationPolicy;
    1618using EpsGreedyPolicy = HeuristicLab.Algorithms.Bandits.BanditPolicies.EpsGreedyPolicy;
    … …  
    2527
    2628      //RunDemo();
    27       //RunGpDemo();
     29      RunGpDemo();
    2830      // RunGridTest();
    29       RunGpGridTest();
     31      //RunGpGridTest();
    3032    }
    3133
    … …  
    297299      const int maxIterations = 100000;
    298300
     301      //var prob = new SymbolicRegressionProblem(new Random(31415), "Tower");
     302      var prob = new SymbolicRegressionPoly10Problem();
     303      var sgp = new OffspringSelectionGP(prob, new Random(seed), true);
     304      RunGP(sgp, prob, 200000, 500, 0.15, 50);
    299305    }
    300306
    … …  
    303309      const int nReps = 20;
    304310      const int seed = 31415;
    305       const int maxIters = 100000;
     311      const int maxIters = 200000;
    306312      var rand = new Random(seed);
    307313      var problemFactories = new Func<ISymbolicExpressionTreeProblem>[]
    308314      {
     315        () => new SymbolicRegressionPoly10Problem(),
    309316        () => new SantaFeAntProblem(),
    310         () => new SymbolicRegressionPoly10Problem(),
    311       };
    312       foreach (var popSize in new int[] { 50, 100, 250, 500, 1000, 2500, 5000 }) {
    313         foreach (var mutationRate in new double[] {/* 0.05, 0.10, */ 0.15, /* 0.25, 0.3 */ }) {
    314           foreach (var maxSize in new int[] { 30, 50, 100 }) {
    315             foreach (var problemFactory in problemFactories)
    316               for (int i = 0; i < nReps; i++) {
    317                 var solverSeed = rand.Next();
    318                 {
    319                   var prob = problemFactory();
    320                   RunStandardGP(prob, solverSeed, maxIters, popSize, mutationRate, maxSize);
     317      };
     318      foreach (var popSize in new int[] { 50, 100, 250, 500, 1000, 2500, 5000, 10000 }) {
     319        foreach (var mutationRate in new double[] { /* 0.05, /* 0.10, */ 0.15, /* 0.25, 0.3 */}) {
     320          foreach (var maxSize in new int[] { 30, 50, 100, 150, 250 }) {
     321            // skip experiments that are already done
     322            if (popSize == 10000 || maxSize == 150 || maxSize == 250) {
     323              foreach (var problemFactory in problemFactories)
     324                for (int i = 0; i < nReps; i++) {
     325                  var solverSeed = rand.Next();
     326                  {
     327                    var prob = problemFactory();
     328                    var sgp = new StandardGP(prob, new Random(solverSeed));
     329                    RunGP(sgp, prob, maxIters, popSize, mutationRate, maxSize);
     330                  }
     331                  // {
     332                  //   var prob = problemFactory();
     333                  //   var osgp = new OffspringSelectionGP(prob, new Random(solverSeed));
     334                  //   RunGP(osgp, prob, maxIters, popSize, mutationRate, maxSize);
     335                  // }
    321336                }
    322                 {
    323                   var prob = problemFactory();
    324                   RunOSGP(prob, solverSeed, maxIters, popSize, mutationRate, maxSize);
    325                 }
    326               }
     337            }
    327338          }
    328339        }
    … …  
    330341    }
    331342
    332     private static void RunStandardGP(ISymbolicExpressionTreeProblem prob, int solverSeed, int maxIters, int popSize, double mutationRate, int maxSize) {
     343    private static void RunGP(IGPSolver gp, ISymbolicExpressionTreeProblem prob, int maxIters, int popSize, double mutationRate, int maxSize) {
    333344      int iterations = 0;
    334345      var globalStatistics = new SentenceSetStatistics(prob.BestKnownQuality(maxSize));
    335 
    336       var gp = new StandardGP(prob, new Random(solverSeed));
     346      var gpName = gp.GetType().Name;
     347      var probName = prob.GetType().Name;
    337348      gp.SolutionEvaluated += (sentence, quality) => {
    338349        iterations++;
    339350        globalStatistics.AddSentence(sentence, quality);
    340351
    341         if (iterations % 10000 == 0) {
    342           Console.WriteLine("\"{0,25}\" \"{1,25}\" {2}", gp, prob, globalStatistics);
     352        if (iterations % 1000 == 0) {
     353          Console.WriteLine("\"{0,25}\" {1} {2:N2} {3} \"{4,25}\" {5}", gpName, popSize, mutationRate, maxSize, probName, globalStatistics);
    343354        }
    344355      };
    … …  
    349360      gp.MaxSolutionDepth = maxSize + 2;
    350361
    351       var sw = new Stopwatch();
    352 
    353       sw.Start();
    354362      gp.Run(maxIters);
    355       sw.Stop();
    356 
    357       Console.WriteLine("\"{0,25}\" \"{1,25}\" {2}", gp, prob, globalStatistics);
    358 
    359       // Console.WriteLine("{0:F2} sec {1,10:F1} sols/sec {2,10:F1} ns/sol",
    360       //   sw.Elapsed.TotalSeconds,
    361       //   maxIters / (double)sw.Elapsed.TotalSeconds,
    362       //   (double)sw.ElapsedMilliseconds * 1000 / maxIters);
    363     }
    364 
    365     private static void RunOSGP(ISymbolicExpressionTreeProblem prob, int solverSeed, int maxIters, int popSize, double mutationRate, int maxSize) {
    366       int iterations = 0;
    367       var globalStatistics = new SentenceSetStatistics(prob.BestKnownQuality(maxSize));
    368 
    369       var gp = new OffspringSelectionGP(prob, new Random(solverSeed));
    370       gp.SolutionEvaluated += (sentence, quality) => {
    371         iterations++;
    372         globalStatistics.AddSentence(sentence, quality);
    373 
    374         if (iterations % 10000 == 0) {
    375           Console.WriteLine("\"{0,25}\" \"{1,25}\" {2}", gp, prob, globalStatistics);
    376         }
    377       };
    378 
    379       gp.PopulationSize = popSize;
    380       gp.MutationRate = mutationRate;
    381       gp.MaxSolutionSize = maxSize + 2;
    382       gp.MaxSolutionDepth = maxSize + 2;
    383 
    384       var sw = new Stopwatch();
    385 
    386       sw.Start();
    387       gp.Run(maxIters);
    388       sw.Stop();
    389 
    390       Console.WriteLine("\"{0,25}\" \"{1,25}\" {2}", gp, prob, globalStatistics);
    391 
    392       // Console.WriteLine("{0:F2} sec {1,10:F1} sols/sec {2,10:F1} ns/sol",
    393       //   sw.Elapsed.TotalSeconds,
    394       //   maxIters / (double)sw.Elapsed.TotalSeconds,
    395       //   (double)sw.ElapsedMilliseconds * 1000 / maxIters);
    396363    }
    397364  }

  • branches/HeuristicLab.Problems.GrammaticalOptimization/Test/TestSolvers.cs

    r11851 r11895  
    22using HeuristicLab.Algorithms.GeneticProgramming;
    33using HeuristicLab.Algorithms.GrammaticalOptimization;
     4using HeuristicLab.Problems.GrammaticalOptimization.SymbReg;
    45using Microsoft.VisualStudio.TestTools.UnitTesting;
    56
    … …  
    154155
    155156    [TestMethod]
    156     public void TestOSGP() {
    157       var prob = new SymbolicRegressionPoly10Problem();
     157    public void TestOSGP()
     158    {
     159      const int seed = 31415;
     160      var prob = new SymbolicRegressionProblem(new Random(seed), "Tower");
    158161      var rand = new Random(31415);
    159162      var osgp = new OffspringSelectionGP(prob, rand);
    160       osgp.PopulationSize = 1000;
    161       osgp.MaxSolutionSize = 50;
     163      osgp.PopulationSize = 500;
     164      osgp.MaxSolutionSize = 100;
    162165      osgp.MaxSolutionDepth = 20;
    163166
    … …  
    166169
    167170      osgp.FoundNewBestSolution += (sentence, quality) => {
    168         Assert.Inconclusive(string.Format("{0:N3} {1}", quality, sentence));
     171        // Assert.Inconclusive(string.Format("{0:N3} {1}", quality, sentence));
    169172        bestSentence = sentence;
    170173        bestQuality = quality;

  • branches/HeuristicLab.Problems.GrammaticalOptimization/Test/TestSymbRegInstances.cs

    r11832 r11895  
    11using System;
    22using System.Linq;
     3using HeuristicLab.Algorithms.GeneticProgramming;
    34using HeuristicLab.Algorithms.GrammaticalOptimization;
    45using HeuristicLab.Problems.GrammaticalOptimization.SymbReg;
    … …  
    1314
    1415      var problem = new SymbolicRegressionProblem(new Random(), "Tower");
    15 
     16      double r2;
     17      Assert.AreEqual(problem.Evaluate("a*b"), problem.OptimizeConstantsAndEvaluate("a*b"));
     18      Assert.AreEqual(problem.OptimizeConstantsAndEvaluate("a*b"), problem.Evaluate("a*b"));
     19      Assert.AreEqual(problem.OptimizeConstantsAndEvaluate("0*a*b"), problem.Evaluate("a*b"), 1E-6);
     20      Assert.AreEqual(problem.OptimizeConstantsAndEvaluate("0*a*b+1"), problem.Evaluate("a*b"), 1E-6);
     21      Assert.IsTrue(problem.OptimizeConstantsAndEvaluate("0*a+b") >= problem.Evaluate("a+b"));
     22      Assert.AreEqual(problem.OptimizeConstantsAndEvaluate("0*a+0*b"), problem.Evaluate("a+b"), 1E-6);
     23      Assert.IsTrue(problem.OptimizeConstantsAndEvaluate("0*a+1*b") >= problem.Evaluate("a+b"));
    1624    }
    1725  }
Note: See TracChangeset for help on using the changeset viewer.