/// <summary>
/// Construct the LMA object.
/// </summary>
/// <param name="network">The network to train. Must have a single output neuron.</param>
/// <param name="training">The training data to use. Must be indexable.</param>
/// <param name="h">The Hessian calculator to use.</param>
public LevenbergMarquardtTraining(BasicNetwork network,
IMLDataSet training, IComputeHessian h)
: base(TrainingImplementationType.Iterative)
{
ValidateNetwork.ValidateMethodToData(network, training);
Training = training;
_indexableTraining = Training;
this._network = network;
_trainingLength = (int)_indexableTraining.Count;
_weightCount = this._network.Structure.CalculateSize();
_lambda = 0.1;
_deltas = new double[_weightCount];
_diagonal = new double[_weightCount];
var input = new BasicMLData(
_indexableTraining.InputSize);
var ideal = new BasicMLData(
_indexableTraining.IdealSize);
_pair = new BasicMLDataPair(input, ideal);
_hessian = h;
_hessian.Init(network, training);
}
/// <inheritdoc />
public override void Iteration()
{
if (!_initComplete)
{
_hessian.Init(_network, Training);
_initComplete = true;
}
PreIteration();
_hessian.Clear();
_weights = NetworkCODEC.NetworkToArray(_network);
_hessian.Compute();
double currentError = _hessian.SSE;
SaveDiagonal();
double startingError = currentError;
bool done = false;
bool singular;
while (!done)
{
ApplyLambda();
var decomposition = new LUDecomposition(_hessian.HessianMatrix);
singular = decomposition.IsNonsingular;
if (singular)
{
_deltas = decomposition.Solve(_hessian.Gradients);
UpdateWeights();
currentError = CalculateError();
}
if (!singular || currentError >= startingError)
{
_lambda *= ScaleLambda;
if (_lambda > LambdaMax)
{
_lambda = LambdaMax;
done = true;
}
}
else
{
_lambda /= ScaleLambda;
done = true;
}
}
Error = currentError;
PostIteration();
}
/// <summary>
/// Construct the LMA object.
/// </summary>
/// <param name="network">The network to train. Must have a single output neuron.</param>
/// <param name="training">The training data to use. Must be indexable.</param>
/// <param name="h">The Hessian calculator to use.</param>
public LevenbergMarquardtTraining(BasicNetwork network,
IMLDataSet training, IComputeHessian h)
: base(TrainingImplementationType.Iterative)
{
ValidateNetwork.ValidateMethodToData(network, training);
Training = training;
_indexableTraining = Training;
this._network = network;
_trainingLength = (int) _indexableTraining.Count;
_weightCount = this._network.Structure.CalculateSize();
_lambda = 0.1;
_deltas = new double[_weightCount];
_diagonal = new double[_weightCount];
var input = new BasicMLData(
_indexableTraining.InputSize);
var ideal = new BasicMLData(
_indexableTraining.IdealSize);
_hessian = h;
_hessian.Init(network, training);
}