/// <summary>
/// Construct restricted Boltzmann machine.
/// </summary>
/// <param name="theLayer">The layer that this RBM works with.</param>
public RestrictedBoltzmannMachine(HiddenLayer theLayer)
{
_layer = theLayer;
_owner = theLayer.Owner;
_hBias = _layer.Bias;
_vBias = new double[VisibleCount];
}
/// <summary>
/// Construct restricted Boltzmann machine.
/// </summary>
/// <param name="theLayer">The layer that this RBM works with.</param>
public RestrictedBoltzmannMachine(HiddenLayer theLayer)
{
_layer = theLayer;
_owner = theLayer.Owner;
_hBias = _layer.Bias;
_vBias = new double[VisibleCount];
}
/// <summary>
/// Construct the supervised trainer for DBN.
/// </summary>
/// <param name="theNetwork">The network to train.</param>
/// <param name="theTrainingInput">The input (x) to train.</param>
/// <param name="theTrainingIdeal">The expected output (y, or labels) to train.</param>
/// <param name="theLearningRate">The learning rate.</param>
public SupervisedTrainDBN(DeepBeliefNetwork theNetwork, double[][] theTrainingInput, double[][] theTrainingIdeal,
double theLearningRate)
{
_network = theNetwork;
_trainingInput = theTrainingInput;
_learningRate = theLearningRate;
_trainingIdeal = theTrainingIdeal;
ErrorCalc = new ErrorCalculationMSE();
}
/// <summary>
/// Construct a trainer for upsupervised training for the DBNN.
/// </summary>
/// <param name="theNetwork">The DBNN to train.</param>
/// <param name="theLevel">The level of the DBNN being trained.</param>
/// <param name="theTrainingInput">The training input cases.</param>
/// <param name="theLearningRate">The learning rate.</param>
/// <param name="theK">The number of cycles to use per iteration.</param>
public UnsupervisedTrainDBN(DeepBeliefNetwork theNetwork, int theLevel, double[][] theTrainingInput,
double theLearningRate, int theK)
{
_network = theNetwork;
_level = theLevel;
_trainingInput = theTrainingInput;
_learningRate = theLearningRate;
_k = theK;
}
/// <summary>
/// Construct a trainer for upsupervised training for the DBNN.
/// </summary>
/// <param name="theNetwork">The DBNN to train.</param>
/// <param name="theLevel">The level of the DBNN being trained.</param>
/// <param name="theTrainingInput">The training input cases.</param>
/// <param name="theLearningRate">The learning rate.</param>
/// <param name="theK">The number of cycles to use per iteration.</param>
public UnsupervisedTrainDBN(DeepBeliefNetwork theNetwork, int theLevel, double[][] theTrainingInput,
double theLearningRate, int theK)
{
_network = theNetwork;
_level = theLevel;
_trainingInput = theTrainingInput;
_learningRate = theLearningRate;
_k = theK;
}
/// <summary>
/// Construct the supervised trainer for DBN.
/// </summary>
/// <param name="theNetwork">The network to train.</param>
/// <param name="theTrainingInput">The input (x) to train.</param>
/// <param name="theTrainingIdeal">The expected output (y, or labels) to train.</param>
/// <param name="theLearningRate">The learning rate.</param>
public SupervisedTrainDBN(DeepBeliefNetwork theNetwork, double[][] theTrainingInput, double[][] theTrainingIdeal,
double theLearningRate)
{
_network = theNetwork;
_trainingInput = theTrainingInput;
_learningRate = theLearningRate;
_trainingIdeal = theTrainingIdeal;
ErrorCalc = new ErrorCalculationMSE();
}
/// <summary>
/// The entry point for this example. If you would like to make this example
/// stand alone, then add to its own project and rename to Main.
/// </summary>
/// <param name="args">Not used.</param>
public static void ExampleMain(string[] args)
{
{
// Create an dbnn belief network.
int[] hidden = { 2, 3 };
DeepBeliefNetwork dbn = new DeepBeliefNetwork(TRAINING_INPUT[0].Length, hidden, TRAINING_IDEAL[0].Length);
dbn.Random = new MersenneTwisterGenerateRandom(54321);
dbn.Reset();
// Layer by layer unsupervised training.
for (int level = 0; level < hidden.Length; level++)
{
UnsupervisedTrainDBN trainUnsupervised = new UnsupervisedTrainDBN(
dbn, level, TRAINING_INPUT, LearningRateUnsupervised, K);
for (int i = 0; i < 2000; i++)
{
trainUnsupervised.Iteration();
}
}
// Supervised training.
SupervisedTrainDBN trainSupervised = new SupervisedTrainDBN(
dbn, TRAINING_INPUT, TRAINING_IDEAL, LearningRateSupervised);
int iteration = 0;
do
{
iteration++;
trainSupervised.Iteration();
Console.WriteLine("Iteration: " + iteration + ", Supervised training: error = "
+ trainSupervised.LastError);
} while (trainSupervised.LastError > 0.001);
// Use test data.
foreach (double[] input in TEST_INPUT)
{
double[] output = dbn.ComputeRegression(input);
Console.WriteLine(string.Join(",",input) + " -> " + string.Join(",",output));
}
}
}
/// <summary>
/// Construct the layer.
/// </summary>
/// <param name="theOwner">The network that owns this layer.</param>
/// <param name="inputCount">The input count for this layer.</param>
/// <param name="outputCount">The output count for this layer.</param>
public DeepLayer(DeepBeliefNetwork theOwner, int inputCount, int outputCount)
{
_weights = AIFH.Alloc2D <double>(outputCount, inputCount);
_bias = new double[outputCount];
_owner = theOwner;
}
/// <summary>
/// Create a hidden layer for a DBNN.
/// </summary>
/// <param name="theOwner">The DBNN that this layer belongs to.</param>
/// <param name="theInputCount">The number of visible units, the input.</param>
/// <param name="theOutputCount">The number of hidden units, the output.</param>
public HiddenLayer(DeepBeliefNetwork theOwner, int theInputCount, int theOutputCount) :
base(theOwner, theInputCount, theOutputCount)
{
}
/// <summary>
/// Create a hidden layer for a DBNN.
/// </summary>
/// <param name="theOwner">The DBNN that this layer belongs to.</param>
/// <param name="theInputCount">The number of visible units, the input.</param>
/// <param name="theOutputCount">The number of hidden units, the output.</param>
public HiddenLayer(DeepBeliefNetwork theOwner, int theInputCount, int theOutputCount) :
base(theOwner, theInputCount, theOutputCount)
{
}
/// <summary>
/// Construct the layer.
/// </summary>
/// <param name="theOwner">The network that owns this layer.</param>
/// <param name="inputCount">The input count for this layer.</param>
/// <param name="outputCount">The output count for this layer.</param>
public DeepLayer(DeepBeliefNetwork theOwner, int inputCount, int outputCount)
{
_weights = AIFH.Alloc2D<double>(outputCount, inputCount);
_bias = new double[outputCount];
_owner = theOwner;
}