source: branches/2789_MathNetNumerics-Exploration/HeuristicLab.Algorithms.DataAnalysis.Experimental/ForwardSelection.cs @ 17607

#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 System.Threading;
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 HeuristicLab.Problems.DataAnalysis;

namespace HeuristicLab.Algorithms.DataAnalysis.Experimental {
  /// <summary>
  /// Forward selection meta-algorithm.
  /// </summary>
  [Item("Forward Selection", "Meta-algorithm that performs feature selection for a given base algorithm using greedy forward selection.")]
  [Creatable(CreatableAttribute.Categories.DataAnalysisRegression, Priority = 999)]
  [StorableClass]
  public sealed class ForwardsSelectionAlgorithm : FixedDataAnalysisAlgorithm<IRegressionProblem> {
    public IFixedValueParameter<IntValue> MaximumInputsParameter {
      get { return (IFixedValueParameter<IntValue>)Parameters["Maximum Inputs"]; }
    }
    public int MaximumInputs {
      get { return MaximumInputsParameter.Value.Value; }
      set { MaximumInputsParameter.Value.Value = value; }
    }

    public IValueParameter<FixedDataAnalysisAlgorithm<IRegressionProblem>> AlgorithmParameter {
      get { return (IValueParameter<FixedDataAnalysisAlgorithm<IRegressionProblem>>)Parameters["Algorithm"]; }
    }

    public FixedDataAnalysisAlgorithm<IRegressionProblem> Algorithm {
      get { return AlgorithmParameter.Value; }
      set { AlgorithmParameter.Value = value; }
    }


    [StorableConstructor]
    private ForwardsSelectionAlgorithm(bool deserializing) : base(deserializing) { }
    [StorableHook(HookType.AfterDeserialization)]
    private void AfterDeserialization() {
      RegisterEventHandlers();
    }

    private ForwardsSelectionAlgorithm(ForwardsSelectionAlgorithm original, Cloner cloner)
      : base(original, cloner) {
      RegisterEventHandlers();
    }
    public override IDeepCloneable Clone(Cloner cloner) {
      return new ForwardsSelectionAlgorithm(this, cloner);
    }

    public ForwardsSelectionAlgorithm()
      : base() {
      Parameters.Add(new FixedValueParameter<IntValue>("Maximum Inputs", "The maximum number of input variables used in the models.", new IntValue(1)));
      Parameters.Add(new ValueParameter<FixedDataAnalysisAlgorithm<IRegressionProblem>>("Algorithm", "The base algorithm for modeling", new LinearRegression()));

      Problem = new RegressionProblem();
      RegisterEventHandlers();
    }

    private void RegisterEventHandlers() {
      Problem.ProblemDataChanged += (o, e) => { MaximumInputs = Problem.ProblemData.InputVariables.CheckedItems
        .Select(t => t.Value)
        .Where(v => Problem.ProblemData.Dataset.VariableHasType<double>(v.Value))
        .Count();
      };
    }
    protected override void OnProblemChanged() {
      base.OnProblemChanged();
      MaximumInputs = Problem.ProblemData.InputVariables.CheckedItems
        .Select(t => t.Value)
        .Where(v => Problem.ProblemData.Dataset.VariableHasType<double>(v.Value))
        .Count();
    }

    protected override void Run(CancellationToken cancellationToken) {
      InitResults();
      var problemClone = (IRegressionProblem)Problem.Clone();
      var problemDataClone = (IRegressionProblemData)problemClone.ProblemData;
      var allowedInputVariables = problemDataClone.InputVariables.CheckedItems.Select(t=>t.Value)
        .Where(v => problemDataClone.Dataset.VariableHasType<double>(v.Value))
        .ToList();
      foreach (var variable in problemDataClone.InputVariables)
        problemDataClone.InputVariables.SetItemCheckedState(variable, false);

      var alg = Algorithm;
      alg.Problem = problemClone;

      for (int inputs = 1; inputs <= MaximumInputs; inputs++) {
        var bestRMSE = double.MaxValue;
        IRegressionSolution bestSolution = null;
        StringValue bestInput = null;
        foreach (var inputVar in allowedInputVariables) {
          if (cancellationToken.IsCancellationRequested) {
            cancellationToken.ThrowIfCancellationRequested();
          }

          problemDataClone.InputVariables.SetItemCheckedState(inputVar, true);
          var solution = RunAlg(alg);
          if (solution != null && solution.TrainingRootMeanSquaredError < bestRMSE) {
            bestRMSE = solution.TrainingRootMeanSquaredError;
            bestSolution = solution;
            bestInput = inputVar;
          }
          problemDataClone.InputVariables.SetItemCheckedState(inputVar, false);
        }

        allowedInputVariables.Remove(bestInput);
        problemDataClone.InputVariables.SetItemCheckedState(bestInput, true);

        bestSolution.Name = inputs.ToString() + " " + bestSolution.Name;
        Results["Current solution"].Value = bestSolution;
        ((ItemList<IRegressionSolution>)Results["All Solutions"].Value).Add(bestSolution);
        ((IntValue)Results["Number of variables"].Value).Value = inputs;
        ((DataTable)Results["RMSE table"].Value).Rows["RMSE (train)"].Values.Add(bestSolution.TrainingRootMeanSquaredError);
        ((DataTable)Results["RMSE table"].Value).Rows["RMSE (test)"].Values.Add(bestSolution.TestRootMeanSquaredError);

        if (cancellationToken.IsCancellationRequested) {
          cancellationToken.ThrowIfCancellationRequested();
        }
      }
    }

    private void InitResults() {
      Results.Add(new Result("Current solution", typeof(IRegressionSolution)));
      Results.Add(new Result("All Solutions", new ItemList<IRegressionSolution>()));
      Results.Add(new Result("Number of variables", new IntValue(0)));
      var rmseTable = new DataTable("RMSE table");
      var trainingRmseRow = new DataRow("RMSE (train)");
      var testRmseRow = new DataRow("RMSE (test)");
      rmseTable.Rows.Add(trainingRmseRow);
      rmseTable.Rows.Add(testRmseRow);
      Results.Add(new Result("RMSE table", rmseTable));
    }

    private IRegressionSolution RunAlg(FixedDataAnalysisAlgorithm<IRegressionProblem> alg) {
      using (var wh = new AutoResetEvent(false)) {
        EventHandler<EventArgs<Exception>> setWhForException = (sender, args) => { wh.Set(); };
        EventHandler setWh = (sender, args) => { wh.Set(); };
        try {
          alg.ExceptionOccurred += setWhForException;
          alg.Stopped += setWh;
          alg.Prepare(true);
          alg.Start();

          wh.WaitOne();

          return alg.Results.Select(r => r.Value).OfType<IRegressionSolution>().FirstOrDefault();
        } finally {
          alg.ExceptionOccurred -= setWhForException;
          alg.Stopped -= setWh;
        }
      }
    }
  }
}