#region License Information
/* HeuristicLab
* Copyright (C) 2002-2015 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 .
*/
#endregion
using System.Linq;
using HeuristicLab.Common;
using HeuristicLab.Core;
using HeuristicLab.Data;
using HeuristicLab.Encodings.RealVectorEncoding;
using HeuristicLab.Operators;
using HeuristicLab.Parameters;
using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
namespace HeuristicLab.Algorithms.GradientDescent {
[StorableClass]
[Item(Name = "LBFGS UpdateResults", Description = "Sets the results (function value and gradients) for the next optimization step in the LM-BFGS algorithm.")]
public sealed class LbfgsUpdateResults : SingleSuccessorOperator {
private const string QualityGradientsParameterName = "QualityGradients";
private const string QualityParameterName = "Quality";
private const string StateParameterName = "State";
private const string ApproximateGradientsParameterName = "ApproximateGradients";
private const string MaximizationParameterName = "Maximization";
#region Parameter Properties
public ILookupParameter ApproximateGradientsParameter {
get { return (ILookupParameter)Parameters[ApproximateGradientsParameterName]; }
}
public ILookupParameter QualityGradientsParameter {
get { return (ILookupParameter)Parameters[QualityGradientsParameterName]; }
}
public ILookupParameter QualityParameter {
get { return (ILookupParameter)Parameters[QualityParameterName]; }
}
public ILookupParameter StateParameter {
get { return (ILookupParameter)Parameters[StateParameterName]; }
}
public ILookupParameter MaximizationParameter {
get { return (ILookupParameter)Parameters[MaximizationParameterName]; }
}
#endregion
#region Properties
private BoolValue ApproximateGradients { get { return ApproximateGradientsParameter.ActualValue; } }
private RealVector QualityGradients { get { return QualityGradientsParameter.ActualValue; } }
private DoubleValue Quality { get { return QualityParameter.ActualValue; } }
private LbfgsState State { get { return StateParameter.ActualValue; } }
private BoolValue Maximization {
get {
// BackwardsCompatibility3.3
#region Backwards compatible code, remove with 3.4
// the parameter is new, previously we assumed minimization problems
if (MaximizationParameter.ActualValue == null) return new BoolValue(false);
#endregion
return MaximizationParameter.ActualValue;
}
}
#endregion
[StorableConstructor]
private LbfgsUpdateResults(bool deserializing) : base(deserializing) { }
private LbfgsUpdateResults(LbfgsUpdateResults original, Cloner cloner) : base(original, cloner) { }
public LbfgsUpdateResults()
: base() {
// in
Parameters.Add(new LookupParameter(QualityGradientsParameterName, "The gradients at the evaluated point of the function to optimize."));
Parameters.Add(new LookupParameter(QualityParameterName, "The value at the evaluated point of the function to optimize."));
Parameters.Add(new LookupParameter(ApproximateGradientsParameterName,
"Flag that indicates if gradients should be approximated."));
Parameters.Add(new LookupParameter(MaximizationParameterName, "Flag that indicates if we solve a maximization problem."));
// in & out
Parameters.Add(new LookupParameter(StateParameterName, "The state of the LM-BFGS algorithm."));
}
[StorableHook(HookType.AfterDeserialization)]
private void AfterDeserialization() {
// BackwardsCompatibility3.3
#region Backwards compatible code, remove with 3.4
if (!Parameters.ContainsKey(MaximizationParameterName)) {
// previous behaviour defaulted to minimization
Parameters.Add(new LookupParameter(MaximizationParameterName, "Flag that indicates if we solve a maximization problem."));
}
#endregion
}
public override IDeepCloneable Clone(Cloner cloner) {
return new LbfgsUpdateResults(this, cloner);
}
public override IOperation Apply() {
var state = State;
var sign = Maximization.Value ? -1.0 : 1.0;
var f = sign * Quality.Value;
state.State.f = f;
if (!ApproximateGradients.Value) {
var g = QualityGradients.Select(gi => sign * gi).ToArray();
state.State.g = g;
}
return base.Apply();
}
}
}