source: branches/PerformanceComparison/HeuristicLab.OptimizationExpertSystem.Common/3.3/Recommenders/DistanceWeightedRecommender.cs @ 13797

#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 HeuristicLab.Analysis;
using HeuristicLab.Common;
using HeuristicLab.Core;
using HeuristicLab.Data;
using HeuristicLab.Optimization;
using HeuristicLab.Parameters;
using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
using System;
using System.Collections.Generic;
using System.Linq;

namespace HeuristicLab.OptimizationExpertSystem.Common {
  [Item("Distance Weighted Recommender", "")]
  [StorableClass]
  public class DistanceWeightedRecommender : ParameterizedNamedItem, IAlgorithmInstanceRecommender {

    private IFixedValueParameter<DoubleValue> NeighborhoodFactorParameter {
      get { return (IFixedValueParameter<DoubleValue>)Parameters["NeighborhoodFactor"]; }
    }

    public double NeighborhoodFactor {
      get { return NeighborhoodFactorParameter.Value.Value; }
      set { NeighborhoodFactorParameter.Value.Value = value; }
    }
    
    [StorableConstructor]
    private DistanceWeightedRecommender(bool deserializing) : base(deserializing) { }
    private DistanceWeightedRecommender(DistanceWeightedRecommender original, Cloner cloner)
      : base(original, cloner) { }
    public DistanceWeightedRecommender() {
      Parameters.Add(new FixedValueParameter<DoubleValue>("NeighborhoodFactor", "Penalize neighbors that are far away.", new DoubleValue(5)));
    }

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

    public IRecommendationModel TrainModel(KnowledgeCenter okc, string[] characteristics) {
      var piDistances = okc.GetProblemDistances(characteristics);
      var maxDist = piDistances.Max(x => x.Value);
      var instances = new SortedList<double, IAlgorithm>();
      foreach (var relevantRuns in okc.GetKnowledgeBaseByAlgorithm()) {
        var algorithm = relevantRuns.Key;
        Func<double, double> distFunc = (d) => Math.Exp(NeighborhoodFactor * (-d / maxDist));
        var pis = relevantRuns.Value.Select(x => ((StringValue)x.Parameters["Problem Name"]).Value).Distinct()
                              .Select(x => Tuple.Create(x, okc.ProblemInstances.SingleOrDefault(y => ((StringValue)y.Parameters["Problem Name"]).Value == x)))
                              .Where(x => x.Item2 != null && x.Item2.Parameters.ContainsKey("BestKnownQuality") && piDistances.ContainsKey(x.Item2))
                              .Select(x => Tuple.Create(x.Item1, distFunc(piDistances[x.Item2]), ((DoubleValue)x.Item2.Parameters["BestKnownQuality"]).Value))
                              .ToDictionary(x => x.Item1, x => Tuple.Create(x.Item2, x.Item3));
        var sumPis = pis.Sum(x => x.Value.Item1);
        var avgERT = 0.0;
        foreach (var problemRuns in relevantRuns.Value.GroupBy(x => ((StringValue)x.Parameters["Problem Name"]).Value)) {
          Tuple<double, double> info;
          if (!pis.TryGetValue(problemRuns.Key, out info)) continue;
          var convGraph = new List<List<Tuple<double, double>>>();
          foreach (var run in problemRuns) {
            var current = new List<Tuple<double, double>>();
            var performanceGraph = ((IndexedDataTable<double>)run.Results["QualityPerEvaluations"]);
            current.AddRange(performanceGraph.Rows.First().Values.TakeWhile(v => v.Item1 < okc.MaximumEvaluations.Value));
            if (current.Count > 0) {
              current.Add(Tuple.Create((double)okc.MaximumEvaluations.Value, current.Last().Item2));
              convGraph.Add(current);
            }
          }
          var ert = ExpectedRuntimeHelper.CalculateErt(convGraph, (okc.Maximization ? (1 - okc.MinimumTarget.Value) : (1 + okc.MinimumTarget.Value)) * info.Item2, okc.Maximization).ExpectedRuntime;
          if (double.IsNaN(ert)) {
            ert = ExpectedRuntimeHelper.CalculateErt(problemRuns.ToList(), "QualityPerEvaluations", (okc.Maximization ? (1 - okc.MinimumTarget.Value) : (1 + okc.MinimumTarget.Value)) * info.Item2, okc.Maximization).ExpectedRuntime;
            if (double.IsNaN(ert)) ert = int.MaxValue;
          }
          avgERT += info.Item1 * ert;
        }
        avgERT /= sumPis;
        if (instances.ContainsKey(avgERT)) {
          avgERT += new System.Random().NextDouble();
        }
        instances.Add(avgERT, (IAlgorithm)algorithm.Clone());
      }

      return new FixedRankModel(instances.Select(x => Tuple.Create(x.Value, x.Key)));
    }
  }
}