source: branches/1614_GeneralizedQAP/HeuristicLab.Analysis.FitnessLandscape/3.3/ProblemCharacteristicAnalysis/GQAP/GQAPDirectedWalk.cs

#region License Information
/* HeuristicLab
 * Copyright (C) 2002-2016 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;
using System.Collections.Generic;
using System.Linq;
using HEAL.Attic;
using HeuristicLab.Common;
using HeuristicLab.Core;
using HeuristicLab.Data;
using HeuristicLab.Encodings.IntegerVectorEncoding;
using HeuristicLab.Optimization;
using HeuristicLab.Parameters;
using HeuristicLab.Problems.GeneralizedQuadraticAssignment;
using HeuristicLab.Random;

namespace HeuristicLab.Analysis.FitnessLandscape {
  [Item("Directed Walk (GQAP-specific)", "")]
  [StorableType("333209A4-8EE7-4944-8A23-CBF120627DBE")]
  public class GQAPDirectedWalk : CharacteristicCalculator {
    
    public IFixedValueParameter<IntValue> PathsParameter {
      get { return (IFixedValueParameter<IntValue>)Parameters["Paths"]; }
    }

    public IFixedValueParameter<BoolValue> BestImprovementParameter {
      get { return (IFixedValueParameter<BoolValue>)Parameters["BestImprovement"]; }
    }

    public IValueParameter<IntValue> SeedParameter {
      get { return (IValueParameter<IntValue>)Parameters["Seed"]; }
    }

    public IFixedValueParameter<BoolValue> LocalOptimaParameter {
      get { return (IFixedValueParameter<BoolValue>)Parameters["LocalOptima"]; }
    }

    public int Paths {
      get { return PathsParameter.Value.Value; }
      set { PathsParameter.Value.Value = value; }
    }

    public bool BestImprovement {
      get { return BestImprovementParameter.Value.Value; }
      set { BestImprovementParameter.Value.Value = value; }
    }

    public int? Seed {
      get { return SeedParameter.Value != null ? SeedParameter.Value.Value : (int?)null; }
      set { SeedParameter.Value = value.HasValue ? new IntValue(value.Value) : null; }
    }

    public bool LocalOptima {
      get { return LocalOptimaParameter.Value.Value; }
      set { LocalOptimaParameter.Value.Value = value; }
    }

    [StorableConstructor]
    private GQAPDirectedWalk(StorableConstructorFlag _) : base(_) { }
    private GQAPDirectedWalk(GQAPDirectedWalk original, Cloner cloner) : base(original, cloner) { }
    public GQAPDirectedWalk() {
      characteristics.AddRange(new[] { "1Shift.Sharpness", "1Shift.Bumpiness", "1Shift.Flatness" }
        .Select(x => new StringValue(x)).ToList());
      Parameters.Add(new FixedValueParameter<IntValue>("Paths", "The number of paths to explore (a path is a set of solutions that connect two randomly chosen solutions).", new IntValue(50)));
      Parameters.Add(new FixedValueParameter<BoolValue>("BestImprovement", "Whether the best of all alternatives should be chosen for each step in the path or just the first improving (least degrading) move should be made.", new BoolValue(true)));
      Parameters.Add(new OptionalValueParameter<IntValue>("Seed", "The seed for the random number generator."));
      Parameters.Add(new FixedValueParameter<BoolValue>("LocalOptima", "Whether to perform walks between local optima.", new BoolValue(false)));
    }

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

    public override bool CanCalculate() {
      return Problem is GQAP;
    }

    public override IEnumerable<IResult> Calculate() {
      IRandom random = Seed.HasValue ? new MersenneTwister((uint)Seed.Value) : new MersenneTwister();
      var gqap = (GQAP)Problem;
      List<IntegerVector> assignments = CalculateRelinkingPoints(random, gqap, Paths, LocalOptima);

      var trajectories = Run(random, (GQAP)Problem, assignments, BestImprovement).ToList();
      var result = IntegerVectorPathAnalysis.GetCharacteristics(trajectories);

      foreach (var chara in characteristics.CheckedItems.Select(x => x.Value.Value)) {
        if (chara == "1Shift.Sharpness") yield return new Result("1Shift.Sharpness", new DoubleValue(result.Sharpness));
        if (chara == "1Shift.Bumpiness") yield return new Result("1Shift.Bumpiness", new DoubleValue(result.Bumpiness));
        if (chara == "1Shift.Flatness") yield return new Result("1Shift.Flatness", new DoubleValue(result.Flatness));
      }
    }

    public static List<IntegerVector> CalculateRelinkingPoints(IRandom random, GQAP gqap, int pathCount, bool localOptima) {
      var assign = new IntegerVector(gqap.ProblemInstance.Demands.Length, random, 0, gqap.ProblemInstance.Capacities.Length);
      if (localOptima) {
        var eval = gqap.ProblemInstance.Evaluate(assign);
        var fit = gqap.ProblemInstance.ToSingleObjective(eval);
        OneOptLocalSearch.Apply(random, assign, ref fit, ref eval, gqap.ProblemInstance, out var evals);
      }
      var points = new List<IntegerVector> { assign };
      while (points.Count < pathCount + 1) {
        assign = (IntegerVector)points.Last().Clone();
        RelocateEquipmentManipluator.Apply(random, assign, gqap.ProblemInstance.Capacities.Length, 0);
        if (localOptima) {
          var eval = gqap.ProblemInstance.Evaluate(assign);
          var fit = gqap.ProblemInstance.ToSingleObjective(eval);
          OneOptLocalSearch.Apply(random, assign, ref fit, ref eval, gqap.ProblemInstance, out var evals);
        }
        if (HammingSimilarityCalculator.CalculateSimilarity(points.Last(), assign) < 0.75)
          points.Add(assign);
      }

      return points;
    }

    public static IEnumerable<List<Tuple<IntegerVector, double>>> Run(IRandom random, GQAP gqap, IEnumerable<IntegerVector> points, bool bestImprovement = true) {
      var iter = points.GetEnumerator();
      if (!iter.MoveNext()) return new List<Tuple<IntegerVector, double>>[0];

      var start = iter.Current;
      var walks = new List<List<Tuple<IntegerVector, double>>>();
      while (iter.MoveNext()) {
        var end = iter.Current;

        var walk = (bestImprovement ? BestDirectedWalk(gqap, start, end) : FirstDirectedWalk(random, gqap, start, end)).ToList();
        walks.Add(walk);
        start = end;
      } // end paths

      var min = walks.SelectMany(x => x.Select(y => y.Item2)).Min();
      var max = walks.SelectMany(x => x.Select(y => y.Item2)).Max();

      if (min == max) max = min + 1;
      return walks.Select(w => w.Select(x => Tuple.Create(x.Item1, (x.Item2 - min) / (max - min))).ToList());
    }

    private static IEnumerable<Tuple<IntegerVector, double>> BestDirectedWalk(GQAP gqap, IntegerVector start, IntegerVector end) {
      var N = gqap.ProblemInstance.Demands.Length;
      var sol = start;
      var evaluation = gqap.ProblemInstance.Evaluate(start);
      var fitness = gqap.ProblemInstance.ToSingleObjective(evaluation);
      yield return Tuple.Create(sol, fitness);
      
      var reassignments = Enumerable.Range(0, N).Select(x => {
        if (start[x] == end[x]) return null;
        var r = new int[N];
        r[x] = end[x] + 1;
        return r;
      }).ToArray();
      var indices = Enumerable.Range(0, N).Select(x => start[x] == end[x] ? null : new List<int>(1) { x }).ToArray();

      for (var step = 0; step < N; step++) {
        var bestFitness = double.MaxValue;
        Evaluation bestEvaluation = null;
        var bestIndex = -1;
        sol = (IntegerVector)sol.Clone();
        
        for (var index = 0; index < N; index++) {
          if (sol[index] == end[index]) continue;

          var oneMove = new NMove(reassignments[index], indices[index]);
          var eval = GQAPNMoveEvaluator.Evaluate(oneMove, sol, evaluation, gqap.ProblemInstance);
          var fit = gqap.ProblemInstance.ToSingleObjective(eval);
          if (fit < bestFitness) { // QAP is minimization
            bestFitness = fit;
            bestEvaluation = eval;
            bestIndex = index;
          }
        }
        if (bestIndex >= 0) {
          sol[bestIndex] = end[bestIndex];
          fitness = bestFitness;
          evaluation = bestEvaluation;
          yield return Tuple.Create(sol, fitness);
        } else break;
      }
    }

    private static IEnumerable<Tuple<IntegerVector, double>> FirstDirectedWalk(IRandom random, GQAP gqap, IntegerVector start, IntegerVector end) {
      var N = gqap.ProblemInstance.Demands.Length;
      var sol = start;
      var evaluation = gqap.ProblemInstance.Evaluate(start);
      var fitness = gqap.ProblemInstance.ToSingleObjective(evaluation);
      yield return Tuple.Create(sol, fitness);

      var reassignments = Enumerable.Range(0, N).Select(x => {
        if (start[x] == end[x]) return null;
        var r = new int[N];
        r[x] = end[x] + 1;
        return r;
      }).ToArray();
      var indices = Enumerable.Range(0, N).Select(x => start[x] == end[x] ? null : new List<int>(1) { x }).ToArray();

      // randomize the order in which improvements are tried
      var order = Enumerable.Range(0, N).Shuffle(random).ToArray();

      for (var step = 0; step < N; step++) {
        var bestFitness = double.MaxValue;
        Evaluation bestEvaluation = null;
        var bestIndex = -1;
        sol = (IntegerVector)sol.Clone();
        for (var i = 0; i < N; i++) {
          var index = order[i];
          if (sol[index] == end[index]) continue;

          var oneMove = new NMove(reassignments[index], indices[index]);
          var eval = GQAPNMoveEvaluator.Evaluate(oneMove, sol, evaluation, gqap.ProblemInstance);
          var fit = gqap.ProblemInstance.ToSingleObjective(eval);

          if (fit < bestFitness) { // GQAP is minimization
            bestFitness = fit;
            bestEvaluation = evaluation;
            bestIndex = index;
            if (fit < fitness) break;
          }
        }
        if (bestIndex >= 0) {
          sol[bestIndex] = end[bestIndex];
          fitness = bestFitness;
          evaluation = bestEvaluation;
          yield return Tuple.Create(sol, fitness);
        } else break;
      }
    }
  }
}