protected NcaGradientCalculator(NcaGradientCalculator original, Cloner cloner) : base(original, cloner) { }
public NcaAlgorithm()
: base() {
Parameters.Add(new ValueParameter<IntValue>(SeedParameterName, "The seed of the random number generator.", new IntValue(0)));
Parameters.Add(new ValueParameter<BoolValue>(SetSeedRandomlyParameterName, "A boolean flag that indicates whether the seed should be randomly reset each time the algorithm is run.", new BoolValue(true)));
Parameters.Add(new FixedValueParameter<IntValue>(KParameterName, "The K for the nearest neighbor.", new IntValue(3)));
Parameters.Add(new FixedValueParameter<IntValue>(DimensionsParameterName, "The number of dimensions that NCA should reduce the data to.", new IntValue(2)));
Parameters.Add(new ConstrainedValueParameter<INcaInitializer>(InitializationParameterName, "Which method should be used to initialize the matrix. Typically LDA (linear discriminant analysis) should provide a good estimate."));
Parameters.Add(new FixedValueParameter<IntValue>(NeighborSamplesParameterName, "How many of the neighbors should be sampled in order to speed up the calculation. This should be at least the value of k and at most the number of training instances minus one will be used.", new IntValue(60)));
Parameters.Add(new FixedValueParameter<IntValue>(IterationsParameterName, "How many iterations the conjugate gradient (CG) method should be allowed to perform. The method might still terminate earlier if a local optima has already been reached.", new IntValue(50)));
Parameters.Add(new FixedValueParameter<DoubleValue>(RegularizationParameterName, "A non-negative paramter which can be set to increase generalization and avoid overfitting. If set to 0 the algorithm is similar to NCA as proposed by Goldberger et al.", new DoubleValue(0)));
Parameters.Add(new ValueParameter<INcaModelCreator>(NcaModelCreatorParameterName, "Creates an NCA model out of the matrix.", new NcaModelCreator()));
Parameters.Add(new ValueParameter<INcaSolutionCreator>(NcaSolutionCreatorParameterName, "Creates an NCA solution given a model and some data.", new NcaSolutionCreator()));
Parameters.Add(new ValueParameter<BoolValue>(ApproximateGradientsParameterName, "True if the gradient should be approximated otherwise they are computed exactly.", new BoolValue()));
NcaSolutionCreatorParameter.Hidden = true;
ApproximateGradientsParameter.Hidden = true;
INcaInitializer defaultInitializer = null;
foreach (var initializer in ApplicationManager.Manager.GetInstances<INcaInitializer>().OrderBy(x => x.ItemName)) {
if (initializer is LdaInitializer) defaultInitializer = initializer;
InitializationParameter.ValidValues.Add(initializer);
}
if (defaultInitializer != null) InitializationParameter.Value = defaultInitializer;
var randomCreator = new RandomCreator();
var ncaInitializer = new Placeholder();
var bfgsInitializer = new LbfgsInitializer();
var makeStep = new LbfgsMakeStep();
var branch = new ConditionalBranch();
var gradientCalculator = new NcaGradientCalculator();
var modelCreator = new Placeholder();
var updateResults = new LbfgsUpdateResults();
var analyzer = new LbfgsAnalyzer();
var finalModelCreator = new Placeholder();
var finalAnalyzer = new LbfgsAnalyzer();
var solutionCreator = new Placeholder();
OperatorGraph.InitialOperator = randomCreator;
randomCreator.SeedParameter.ActualName = SeedParameterName;
randomCreator.SeedParameter.Value = null;
randomCreator.SetSeedRandomlyParameter.ActualName = SetSeedRandomlyParameterName;
randomCreator.SetSeedRandomlyParameter.Value = null;
randomCreator.Successor = ncaInitializer;
ncaInitializer.Name = "(NcaInitializer)";
ncaInitializer.OperatorParameter.ActualName = InitializationParameterName;
ncaInitializer.Successor = bfgsInitializer;
bfgsInitializer.IterationsParameter.ActualName = IterationsParameterName;
bfgsInitializer.PointParameter.ActualName = NcaMatrixParameterName;
bfgsInitializer.ApproximateGradientsParameter.ActualName = ApproximateGradientsParameterName;
bfgsInitializer.Successor = makeStep;
makeStep.StateParameter.ActualName = bfgsInitializer.StateParameter.Name;
makeStep.PointParameter.ActualName = NcaMatrixParameterName;
makeStep.Successor = branch;
branch.ConditionParameter.ActualName = makeStep.TerminationCriterionParameter.Name;
branch.FalseBranch = gradientCalculator;
branch.TrueBranch = finalModelCreator;
gradientCalculator.Successor = modelCreator;
modelCreator.OperatorParameter.ActualName = NcaModelCreatorParameterName;
modelCreator.Successor = updateResults;
updateResults.StateParameter.ActualName = bfgsInitializer.StateParameter.Name;
updateResults.QualityParameter.ActualName = QualityParameterName;
updateResults.QualityGradientsParameter.ActualName = NcaMatrixGradientsParameterName;
updateResults.ApproximateGradientsParameter.ActualName = ApproximateGradientsParameterName;
updateResults.Successor = analyzer;
analyzer.QualityParameter.ActualName = QualityParameterName;
analyzer.PointParameter.ActualName = NcaMatrixParameterName;
analyzer.QualityGradientsParameter.ActualName = NcaMatrixGradientsParameterName;
analyzer.StateParameter.ActualName = bfgsInitializer.StateParameter.Name;
analyzer.PointsTableParameter.ActualName = "Matrix table";
analyzer.QualityGradientsTableParameter.ActualName = "Gradients table";
analyzer.QualitiesTableParameter.ActualName = "Qualities";
analyzer.Successor = makeStep;
finalModelCreator.OperatorParameter.ActualName = NcaModelCreatorParameterName;
finalModelCreator.Successor = finalAnalyzer;
finalAnalyzer.QualityParameter.ActualName = QualityParameterName;
finalAnalyzer.PointParameter.ActualName = NcaMatrixParameterName;
finalAnalyzer.QualityGradientsParameter.ActualName = NcaMatrixGradientsParameterName;
finalAnalyzer.PointsTableParameter.ActualName = analyzer.PointsTableParameter.ActualName;
finalAnalyzer.QualityGradientsTableParameter.ActualName = analyzer.QualityGradientsTableParameter.ActualName;
finalAnalyzer.QualitiesTableParameter.ActualName = analyzer.QualitiesTableParameter.ActualName;
finalAnalyzer.Successor = solutionCreator;
solutionCreator.OperatorParameter.ActualName = NcaSolutionCreatorParameterName;
Problem = new ClassificationProblem();
}
public NcaAlgorithm()
: base()
{
Parameters.Add(new ValueParameter <IntValue>(SeedParameterName, "The seed of the random number generator.", new IntValue(0)));
Parameters.Add(new ValueParameter <BoolValue>(SetSeedRandomlyParameterName, "A boolean flag that indicates whether the seed should be randomly reset each time the algorithm is run.", new BoolValue(true)));
Parameters.Add(new FixedValueParameter <IntValue>(KParameterName, "The K for the nearest neighbor.", new IntValue(3)));
Parameters.Add(new FixedValueParameter <IntValue>(DimensionsParameterName, "The number of dimensions that NCA should reduce the data to.", new IntValue(2)));
Parameters.Add(new ConstrainedValueParameter <INcaInitializer>(InitializationParameterName, "Which method should be used to initialize the matrix. Typically LDA (linear discriminant analysis) should provide a good estimate."));
Parameters.Add(new FixedValueParameter <IntValue>(NeighborSamplesParameterName, "How many of the neighbors should be sampled in order to speed up the calculation. This should be at least the value of k and at most the number of training instances minus one will be used.", new IntValue(60)));
Parameters.Add(new FixedValueParameter <IntValue>(IterationsParameterName, "How many iterations the conjugate gradient (CG) method should be allowed to perform. The method might still terminate earlier if a local optima has already been reached.", new IntValue(50)));
Parameters.Add(new FixedValueParameter <DoubleValue>(RegularizationParameterName, "A non-negative paramter which can be set to increase generalization and avoid overfitting. If set to 0 the algorithm is similar to NCA as proposed by Goldberger et al.", new DoubleValue(0)));
Parameters.Add(new ValueParameter <INcaModelCreator>(NcaModelCreatorParameterName, "Creates an NCA model out of the matrix.", new NcaModelCreator()));
Parameters.Add(new ValueParameter <INcaSolutionCreator>(NcaSolutionCreatorParameterName, "Creates an NCA solution given a model and some data.", new NcaSolutionCreator()));
Parameters.Add(new ValueParameter <BoolValue>(ApproximateGradientsParameterName, "True if the gradient should be approximated otherwise they are computed exactly.", new BoolValue()));
NcaSolutionCreatorParameter.Hidden = true;
ApproximateGradientsParameter.Hidden = true;
INcaInitializer defaultInitializer = null;
foreach (var initializer in ApplicationManager.Manager.GetInstances <INcaInitializer>().OrderBy(x => x.ItemName))
{
if (initializer is LdaInitializer)
{
defaultInitializer = initializer;
}
InitializationParameter.ValidValues.Add(initializer);
}
if (defaultInitializer != null)
{
InitializationParameter.Value = defaultInitializer;
}
var randomCreator = new RandomCreator();
var ncaInitializer = new Placeholder();
var bfgsInitializer = new LbfgsInitializer();
var makeStep = new LbfgsMakeStep();
var branch = new ConditionalBranch();
var gradientCalculator = new NcaGradientCalculator();
var modelCreator = new Placeholder();
var updateResults = new LbfgsUpdateResults();
var analyzer = new LbfgsAnalyzer();
var finalModelCreator = new Placeholder();
var finalAnalyzer = new LbfgsAnalyzer();
var solutionCreator = new Placeholder();
OperatorGraph.InitialOperator = randomCreator;
randomCreator.SeedParameter.ActualName = SeedParameterName;
randomCreator.SeedParameter.Value = null;
randomCreator.SetSeedRandomlyParameter.ActualName = SetSeedRandomlyParameterName;
randomCreator.SetSeedRandomlyParameter.Value = null;
randomCreator.Successor = ncaInitializer;
ncaInitializer.Name = "(NcaInitializer)";
ncaInitializer.OperatorParameter.ActualName = InitializationParameterName;
ncaInitializer.Successor = bfgsInitializer;
bfgsInitializer.IterationsParameter.ActualName = IterationsParameterName;
bfgsInitializer.PointParameter.ActualName = NcaMatrixParameterName;
bfgsInitializer.ApproximateGradientsParameter.ActualName = ApproximateGradientsParameterName;
bfgsInitializer.Successor = makeStep;
makeStep.StateParameter.ActualName = bfgsInitializer.StateParameter.Name;
makeStep.PointParameter.ActualName = NcaMatrixParameterName;
makeStep.Successor = branch;
branch.ConditionParameter.ActualName = makeStep.TerminationCriterionParameter.Name;
branch.FalseBranch = gradientCalculator;
branch.TrueBranch = finalModelCreator;
gradientCalculator.Successor = modelCreator;
modelCreator.OperatorParameter.ActualName = NcaModelCreatorParameterName;
modelCreator.Successor = updateResults;
updateResults.StateParameter.ActualName = bfgsInitializer.StateParameter.Name;
updateResults.QualityParameter.ActualName = QualityParameterName;
updateResults.QualityGradientsParameter.ActualName = NcaMatrixGradientsParameterName;
updateResults.ApproximateGradientsParameter.ActualName = ApproximateGradientsParameterName;
updateResults.Successor = analyzer;
analyzer.QualityParameter.ActualName = QualityParameterName;
analyzer.PointParameter.ActualName = NcaMatrixParameterName;
analyzer.QualityGradientsParameter.ActualName = NcaMatrixGradientsParameterName;
analyzer.StateParameter.ActualName = bfgsInitializer.StateParameter.Name;
analyzer.PointsTableParameter.ActualName = "Matrix table";
analyzer.QualityGradientsTableParameter.ActualName = "Gradients table";
analyzer.QualitiesTableParameter.ActualName = "Qualities";
analyzer.Successor = makeStep;
finalModelCreator.OperatorParameter.ActualName = NcaModelCreatorParameterName;
finalModelCreator.Successor = finalAnalyzer;
finalAnalyzer.QualityParameter.ActualName = QualityParameterName;
finalAnalyzer.PointParameter.ActualName = NcaMatrixParameterName;
finalAnalyzer.QualityGradientsParameter.ActualName = NcaMatrixGradientsParameterName;
finalAnalyzer.PointsTableParameter.ActualName = analyzer.PointsTableParameter.ActualName;
finalAnalyzer.QualityGradientsTableParameter.ActualName = analyzer.QualityGradientsTableParameter.ActualName;
finalAnalyzer.QualitiesTableParameter.ActualName = analyzer.QualitiesTableParameter.ActualName;
finalAnalyzer.Successor = solutionCreator;
solutionCreator.OperatorParameter.ActualName = NcaSolutionCreatorParameterName;
Problem = new ClassificationProblem();
}
protected NcaGradientCalculator(NcaGradientCalculator original, Cloner cloner) : base(original, cloner)
{
}
