source: branches/HeuristicLab.Analysis.AlgorithmBehavior/HeuristicLab.Analysis.AlgorithmBehavior/3.3/Analyzers/BuildingBlockAnalyzer.cs @ 8216

#region License Information
/* HeuristicLab
 * Copyright (C) 2002-2012 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 <http://www.gnu.org/licenses/>.
 */
#endregion

using System.Linq;
using HeuristicLab.Common;
using HeuristicLab.Core;
using HeuristicLab.Operators;
using HeuristicLab.Optimization;
using HeuristicLab.Parameters;
using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;

namespace HeuristicLab.Analysis.AlgorithmBehavior {
  using System;
  using System.Collections.Generic;
  using HeuristicLab.Data;
  using HeuristicLab.Encodings.PermutationEncoding;
  using TraceMapType = HeuristicLab.Core.ItemDictionary<HeuristicLab.Core.IItem, HeuristicLab.Core.IItemList<HeuristicLab.Core.IItem>>;

  /// <summary>
  /// An operator that analyzes building blocks.
  /// </summary>
  [Item("BuildingBlockAnalyzer", "An operator that analyzes building blocks.")]
  [StorableClass]
  public sealed class BuildingBlockAnalyzer : SingleSuccessorOperator, IAnalyzer {
    private const string GlobalTraceMapParameterName = "GlobalTraceMap";
    private const string ResultsParameterName = "Results";
    private const string PopulationGraphResultParameterName = "PopulationGraph";
    private const string BuildingBlockSchemaMatrixParameterName = "BuildingBlockSchemaMatrix";
    private const string BrokenInteritanceSchemaOccurrenceName = "BrokenInteritanceSchemaOccurrenceMatrix";
    private const string UnbrokenInteritanceSchemaOccurrenceName = "UnbrokenInteritanceSchemaOccurrence";
    private const string BuildingBlockPatternMatrixParameterName = "BuildingBlockPatternMatrix";

    #region IAnalyzer Members
    public bool EnabledByDefault {
      get { return true; }
    }
    #endregion

    #region Parameter properties
    public ILookupParameter<TraceMapType> GlobalTraceMapParameter {
      get { return (ILookupParameter<TraceMapType>)Parameters[GlobalTraceMapParameterName]; }
    }
    public ILookupParameter<ResultCollection> ResultsParameter {
      get { return (ILookupParameter<ResultCollection>)Parameters[ResultsParameterName]; }
    }
    #endregion

    #region Properties
    public TraceMapType GlobalTraceMap {
      get { return GlobalTraceMapParameter.ActualValue; }
    }
    public ResultCollection Results {
      get { return ResultsParameter.ActualValue; }
    }
    #endregion

    [StorableConstructor]
    private BuildingBlockAnalyzer(bool deserializing) : base(deserializing) { }
    private BuildingBlockAnalyzer(BuildingBlockAnalyzer original, Cloner cloner) : base(original, cloner) { }
    public BuildingBlockAnalyzer()
      : base() {
      Parameters.Add(new LookupParameter<TraceMapType>(GlobalTraceMapParameterName, "A global cache containing the whole genealogy."));
      Parameters.Add(new LookupParameter<ResultCollection>(ResultsParameterName, "The results collection where the analysis values should are stored."));
    }

    public override IDeepCloneable Clone(Cloner cloner) {
      return new BuildingBlockAnalyzer(this, cloner);
    }

    public override IOperation Apply() {
      var graph = (GenealogyGraph<Permutation>)Results[PopulationGraphResultParameterName].Value;
      var generationZero = graph.Values.Where(x => x.Rank == 0);

      // extract all subtours for each individual in generation zero
      var subtours = new Dictionary<Permutation, IList<int[]>>();
      foreach (var individual in generationZero) {
        subtours[(Permutation)individual.Data] = ExtractSubtours((Permutation)individual.Data);
      }

      // process all schemata
      var schemataMatrix = new Dictionary<int[], int[]>(new SchemaEqualityComparer());
      foreach (var entry in subtours) {
        foreach (var schema in entry.Value) {
          if (!schemataMatrix.ContainsKey(schema)) {
            var categories = new int[3];
            categories[0] = ExploreGlobalSchemaOccurrence(graph, schema);
            categories[1] = -1;
            categories[2] = -1;
            schemataMatrix.Add(schema, categories);
          }
        }
      }

      // order schemas according to their global occurrence count
      var sortedSchemataMatrix = schemataMatrix.OrderByDescending(x => x.Value[0]).ToDictionary(x => x.Key, x => x.Value);
      IList<string> rowNames = new List<string>();
      var schemataStringMatrix = new StringMatrix(sortedSchemataMatrix.Keys.Count, 3);
      for (int i = 0; i < sortedSchemataMatrix.Keys.Count; i++) {
        var element = sortedSchemataMatrix.ElementAt(i);
        rowNames.Add(new Permutation(PermutationTypes.RelativeUndirected, element.Key).ToString());
        for (int j = 0; j < element.Value.Count(); j++) {
          schemataStringMatrix[i, j] = element.Value[j].ToString();
        }
      }
      schemataStringMatrix.RowNames = rowNames;
      schemataStringMatrix.ColumnNames = new[] { "GlobalSchemaOccurrence", "<unused>", "<unused>" };
      Results.Add(new Result(BuildingBlockSchemaMatrixParameterName, schemataStringMatrix));

      var brokenInteritanceSchemaOccurrenceMatrix = new StringMatrix(sortedSchemataMatrix.Keys.Count, generationZero.Count());
      IList<string> brokenRowNames = new List<string>();
      IList<string> brokenColumnNames = new List<string>();
      for (int i = 0; i < sortedSchemataMatrix.Keys.Count; i++) {
        var entry = sortedSchemataMatrix.ElementAt(i);
        brokenRowNames.Add(new Permutation(PermutationTypes.RelativeUndirected, entry.Key).ToString());
        for (int j = 0; j < generationZero.Count(); j++) {
          var individual = generationZero.ElementAt(j);
          if (i == 0) brokenColumnNames.Add(((Permutation)individual.Data).ToString());
          brokenInteritanceSchemaOccurrenceMatrix[i, j] = CalculateBrokenInheritanceOccurrences(entry.Key, individual).ToString();
        }
      }
      brokenInteritanceSchemaOccurrenceMatrix.RowNames = brokenRowNames;
      brokenInteritanceSchemaOccurrenceMatrix.ColumnNames = brokenColumnNames;
      Results.Add(new Result(BrokenInteritanceSchemaOccurrenceName, brokenInteritanceSchemaOccurrenceMatrix));

      var unbrokenInteritanceSchemaOccurrenceMatrix = new StringMatrix(sortedSchemataMatrix.Keys.Count, generationZero.Count());
      IList<string> unbrokenRowNames = new List<string>();
      IList<string> unbrokenColumnNames = new List<string>();
      for (int i = 0; i < sortedSchemataMatrix.Keys.Count; i++) {
        var entry = sortedSchemataMatrix.ElementAt(i);
        unbrokenRowNames.Add(new Permutation(PermutationTypes.RelativeUndirected, entry.Key).ToString());
        for (int j = 0; j < generationZero.Count(); j++) {
          var individual = generationZero.ElementAt(j);
          if (i == 0) unbrokenColumnNames.Add(((Permutation)individual.Data).ToString());
          unbrokenInteritanceSchemaOccurrenceMatrix[i, j] = CalculateUnbrokenInheritanceOccurrences(entry.Key, individual).ToString();
        }
      }
      unbrokenInteritanceSchemaOccurrenceMatrix.RowNames = unbrokenRowNames;
      unbrokenInteritanceSchemaOccurrenceMatrix.ColumnNames = brokenColumnNames;
      Results.Add(new Result(UnbrokenInteritanceSchemaOccurrenceName, unbrokenInteritanceSchemaOccurrenceMatrix));

      return base.Apply();
    }

    private int CalculateBrokenInheritanceOccurrences(int[] schema, GenealogyGraphNode individual) {
      // TODO
      return -1;
    }

    private int CalculateUnbrokenInheritanceOccurrences(int[] schema, GenealogyGraphNode individual) {
      var filteredDescendants = individual.Descendants().Where(x => IsMatch(schema, ((Permutation)x.Data).ToArray()));
      return filteredDescendants.Any() ? (int)filteredDescendants.Max(x => x.Rank) : 0;
    }

    private int ExploreGlobalSchemaOccurrence(GenealogyGraph<Permutation> graph, int[] schema) {
      int occurrences = 0;
      foreach (var individual in graph.Values)
        if (IsSubtour(schema, (Permutation)individual.Data)) occurrences++;
      return occurrences;
    }

    private IList<int[]> ExtractSubtours(Permutation permutation) {
      var subtours = new List<int[]>();
      for (int i = 2; i <= permutation.Count() / 2; i++) { // increase schema length from 2 to n/2
        for (int j = 0; j < permutation.Count(); j++) { // visit all positions in the permutation
          var schema = new int[i];
          for (int k = 0; k < i; k++) schema[k] = permutation[(j + k) % permutation.Length]; // copy edge to schema
          subtours.Add(schema);
        }
      }
      return subtours;
    }

    private bool IsSubtour(int[] tour, Permutation permutation) {
      // determine starting position for subtour match
      int idx = permutation.Select((x, index) => new { Value = x, Index = index }).Single(x => x.Value == tour[0]).Index;
      bool isSubtour = true;
      for (int i = 1; i < tour.Length; i++) {
        if (tour[i] == permutation[(idx + 1) % permutation.Length]) { // check right side
          idx = (idx + 1) % permutation.Length;
        } else if (tour[i] == permutation[(idx - 1 + permutation.Length) % permutation.Length]) { // check left side
          idx = (idx - 1 + permutation.Length) % permutation.Length;
        } else {
          isSubtour = false;
          break;
        }
      }
      return isSubtour;
    }

    // attention: brute force matching
    // should be replaced with KMP, BM or RK
    private bool IsMatch(int[] pattern, int[] tour) {
      bool match = false;
      for (int i = 0; i <= tour.Length - pattern.Length && !match; i++) {
        for (int j = 0; j < pattern.Length && (match = pattern[j] == tour[i + j] || pattern[j] == -1); j++) ;
      }
      return match;
    }

    private IList<int[]> GenerateWildcardSchemata(int[] schema) {
      string formatString = "{0:d" + schema.Length + "}";
      var combs = new List<bool[]>();
      string comb = string.Empty;

      // create all desired wildcard combinations
      int i = 0;
      int wildcardCount = 0; // first wildcard combination contains no wildcards
      while (wildcardCount < schema.Length) {
        if (wildcardCount > schema.Length / 8 && wildcardCount < schema.Length / 2)
          combs.Add(comb.Select(x => x == '1').ToArray()); // '1' becomes true which indicates a wildcard
        wildcardCount = (comb = string.Format(formatString, Convert.ToString(++i, 2))).Count(x => x == '1');
      }

      // create all desired wildcard instances
      var wildcardSchemata = new List<int[]>();
      foreach (var c in combs) {
        wildcardSchemata.Add(schema.Select((x, index) => c[index] ? -1 : x).ToArray()); // -1 is a wildcard in the schema
      }

      return wildcardSchemata;
    }


    private class SchemaEqualityComparer : EqualityComparer<int[]> {
      public override bool Equals(int[] x, int[] y) {
        if (object.ReferenceEquals(x, y)) return true;
        if (x == null || y == null || x.Length != y.Length) return false;
        EqualityComparer<int> comparer = EqualityComparer<int>.Default;
        bool match = true;
        for (int i = 0; i < x.Length && match; i++) // check normal tour
          if (!comparer.Equals(x[i], y[i])) match = false;
        if (!match) {
          match = true;
          for (int i = x.Length - 1; i >= 0 && match; i--) // check inverse tour
            if (!comparer.Equals(x[i], y[i])) match = false;
        }
        return match;
      }

      public override int GetHashCode(int[] obj) {
        return 0; // return the same hash code for each object, otherwise Equals will not be called
      }
    }
  }
}