source: branches/2745_EfficientGlobalOptimization/HeuristicLab.Algorithms.EGO/InfillCriteria/ExpectedQuantileImprovement.cs @ 17357

#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.Linq;
using HEAL.Attic;
using HeuristicLab.Common;
using HeuristicLab.Core;
using HeuristicLab.Data;
using HeuristicLab.Encodings.RealVectorEncoding;
using HeuristicLab.Parameters;
using HeuristicLab.Problems.DataAnalysis;

// ReSharper disable once CheckNamespace
namespace HeuristicLab.Algorithms.EGO {

    [StorableType("9c260dee-1f7c-4c07-bdef-a9e1bf2b26f8")]
    [Item("ExpectedQuantileImprovement", "Noisy InfillCriterion, Extension of the Expected Improvement as described in \n Noisy expectedimprovement and on - line computation time allocation for the optimization of simulators with tunable fidelitys\r\nPicheny, V., Ginsbourger, D., Richet, Y")]
  public class ExpectedQuantileImprovement : ExpectedImprovementBase {

    #region Parameternames
    public const string AlphaParameterName = "Alpha";
    public const string MaxEvaluationsParameterName = "MaxEvaluations";
    #endregion

    #region Parameters
    public IFixedValueParameter<DoubleValue> AlphaParameter => Parameters[AlphaParameterName] as IFixedValueParameter<DoubleValue>;
    public IValueParameter<IntValue> MaxEvaluationsParameter => Parameters[MaxEvaluationsParameterName] as IValueParameter<IntValue>;
    #endregion

    #region Properties
    public int MaxEvaluations => MaxEvaluationsParameter.Value.Value;
    public double Alpha => AlphaParameter.Value.Value;
    [Storable]
    private double Tau;
    #endregion

    #region HL-Constructors, Serialization and Cloning 
    [StorableConstructor]
    protected ExpectedQuantileImprovement(StorableConstructorFlag deserializing) : base(deserializing) { }
    protected ExpectedQuantileImprovement(ExpectedQuantileImprovement original, Cloner cloner) : base(original, cloner) {
      Tau = original.Tau;
    }
    public ExpectedQuantileImprovement() {
      Parameters.Add(new FixedValueParameter<DoubleValue>(AlphaParameterName, "The Alpha value specifiying the robustness of the \"effective best solution\". Recommended value is 1.0", new DoubleValue(1.0)));
      Parameters.Add(new ValueParameter<IntValue>(MaxEvaluationsParameterName, "The maximum number of evaluations allowed for EGO", new IntValue(500)));
      MaxEvaluationsParameter.Hidden = true;
    }
    public override IDeepCloneable Clone(Cloner cloner) {
      return new ExpectedQuantileImprovement(this, cloner);
    }
    #endregion

    protected override double FindBestFitness(IConfidenceRegressionSolution solution) {
      Tau = RegressionSolution.EstimatedTrainingValues.Zip(solution.ProblemData.TargetVariableTrainingValues, (d, d1) => Math.Abs(d - d1)).Average();
      Tau = Tau * Tau / (MaxEvaluations - solution.ProblemData.Dataset.Rows % MaxEvaluations + 1);

      var index = solution.EstimatedTrainingValues.Zip(solution.EstimatedTrainingVariances, (m, s2) => m + Alpha * Math.Sqrt(s2)).ArgMin(x => x);
      return solution.EstimatedTrainingValues.ToArray()[index];

    }

    protected override double Evaluate(RealVector vector, double estimatedFitness, double estimatedStandardDeviation) {
      var s2 = estimatedStandardDeviation * estimatedStandardDeviation;
      var penalty = Alpha * Math.Sqrt(Tau * s2 / (Tau + s2));
      var yhat = estimatedFitness + (ExpensiveMaximization ? -penalty : penalty);
      var s = Math.Sqrt(s2 * s2 / (Tau + s2));
      return GetEstimatedImprovement(BestFitness, yhat, s, ExploitationWeight, ExpensiveMaximization);
    }

  }
}