public BuildGenericBackPropagationRprop()
{
_model = new ModelBackPropagationBase();
_activationFunctionOutputLayer = new Sigmoid();
_factorDecrease = .5;
_factorIncrease = 1.2;
_alpha = .1;
_maxUpdateFactor = Double.MaxValue;
_minUpdateFactor = Double.MinValue;
}
protected bool VerifyUpstreamLayers(ModelBackPropagationBase model)
{
bool exceptionFlag = false;
Parallel.For(1, model.GetNumberOfLayers(),
new ParallelOptions {
MaxDegreeOfParallelism = _maxParallelThreads
},
idxLayer =>
{
if (model.GetLayer(idxLayer).GetNumberOfUpstreamUnits() < 0) //Check if Upstream Layer is setup, this will set an exception otherwise
{
exceptionFlag = true;
}
});
return(exceptionFlag);
}
protected bool VerifyLayers(ModelBackPropagationBase model)
{
bool exceptionFlag = false;
Parallel.For(0, model.GetNumberOfLayers() - 1,
new ParallelOptions {
MaxDegreeOfParallelism = _maxParallelThreads
},
idxLayer =>
{
if (model.GetLayer(idxLayer) == null)
{
exceptionFlag = true;
}
});
return(exceptionFlag);
}
BuildModel(double[][] trainingData,
string[] attributeHeaders,
int indexTargetAttribute)
{
//Verify data and set variables
VerifyData(trainingData, attributeHeaders, indexTargetAttribute);
//model.Mode = _mode; //Classification or Regression
//Find out number of categories
_noOfUnitsOutputLayer = GetNumberOfTargetValues(_mode, trainingData, indexTargetAttribute).Length;
//Set Number of Hidden Layers
if (_noOfUnitsHiddenLayer == int.MinValue)
{
_noOfUnitsHiddenLayer =
GetNumberOfHiddenUnits(_noOfUnitsOutputLayer, _scalingFactor, _noOfAttributes);
}
if (_activationFunctions[0] == null)
{
SetActivationFunction(0, new Sigmoid());
}
if (_activationFunctions[1] == null)
{
SetActivationFunction(1, new Sigmoid());
}
BuildGenericBackPropagationStandard build =
new BuildGenericBackPropagationStandard();
build.SetParameters((double)_mode, 1, _threshold, _noOfEpoch, _alpha);
build.AddHiddenLayer(0, _noOfUnitsHiddenLayer, _activationFunctions[0]);
build.SetOutputLayerActivationFunction(_activationFunctions[1]);
ModelBackPropagationBase model =
(ModelBackPropagationBase)build.BuildModel(
_trainingData, _attributeHeaders,
_indexTargetAttribute);
return(model);
}
public BuildGenericBackPropagationStandard()
{
_model = new ModelBackPropagationBase();
_activationFunctionOutputLayer = new Sigmoid();
}