Package smile.base.mlp

Class MultilayerPerceptron

java.lang.Object
smile.base.mlp.MultilayerPerceptron
All Implemented Interfaces:
Serializable, AutoCloseable
Direct Known Subclasses:
MLP, MLP
public abstract class MultilayerPerceptron extends Object implements AutoCloseable, Serializable
Fully connected multilayer perceptron neural network. An MLP consists of at least three layers of nodes: an input layer, a hidden layer and an output layer. The nodes are interconnected through weighted acyclic arcs from each preceding layer to the following, without lateral or feedback connections. Each node calculates a transformed weighted linear combination of its inputs (output activations from the preceding layer), with one of the weights acting as a trainable bias connected to a constant input. The transformation, called activation function, is a bounded non-decreasing (non-linear) function.
See Also:

  • Field Summary

    Fields
    Modifier and Type
    Field
    Description
    protected double
    clipNorm
    The gradient clipping norm.
    protected double
    clipValue
    The gradient clipping value.
    protected double
    epsilon
    A small constant for numerical stability in RMSProp.
    protected double
    lambda
    The L2 regularization factor, which is also the weight decay factor.
    protected TimeFunction
    learningRate
    The learning rate.
    protected TimeFunction
    momentum
    The momentum factor.
    protected Layer[]
    net
    The input and hidden layers.
    protected OutputLayer
    output
    The output layer.
    protected int
    p
    The dimensionality of input data.
    protected double
    rho
    The discounting factor for the history/coming gradient in RMSProp.
    protected int
    t
    The training iterations.
    protected ThreadLocal<double[]>
    target
    The buffer to store desired target value of training instance.

  • Constructor Summary

    Constructors
    Constructor
    Description
    MultilayerPerceptron(Layer... net)
    Constructor.

  • Method Summary

    Modifier and Type
    Method
    Description
    protected void
    backpropagate(boolean update)
    Propagates the errors back through the network.
    void
    close()
     
    double
    getClipNorm()
    Returns the gradient clipping norm.
    double
    getClipValue()
    Returns the gradient clipping value.
    double
    getLearningRate()
    Returns the learning rate.
    double
    getMomentum()
    Returns the momentum factor.
    double
    getWeightDecay()
    Returns the weight decay factor.
    protected void
    propagate(double[] x, boolean training)
    Propagates the signals through the neural network.
    void
    setClipNorm(double clipNorm)
    Sets the gradient clipping norm.
    void
    setClipValue(double clipValue)
    Sets the gradient clipping value.
    void
    setLearningRate(TimeFunction rate)
    Sets the learning rate.
    void
    setMomentum(TimeFunction momentum)
    Sets the momentum factor.
    void
    setParameters(Properties params)
    Sets MLP hyperparameters such as learning rate, weight decay, momentum, RMSProp, etc.
    void
    setRMSProp(double rho, double epsilon)
    Sets RMSProp parameters.
    void
    setWeightDecay(double lambda)
    Sets the weight decay factor.
    String
    toString()
     
    protected void
    update(int m)
    Updates the weights for mini-batch training.

    Methods inherited from class java.lang.Object

    clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

  • Field Details

    • p

      protected int p
      The dimensionality of input data.

    • output

      protected OutputLayer output
      The output layer.

    • net

      protected Layer[] net
      The input and hidden layers.

    • target

      protected transient ThreadLocal<double[]> target
      The buffer to store desired target value of training instance.

    • learningRate

      protected TimeFunction learningRate
      The learning rate.

    • momentum

      protected TimeFunction momentum
      The momentum factor.

    • rho

      protected double rho
      The discounting factor for the history/coming gradient in RMSProp.

    • epsilon

      protected double epsilon
      A small constant for numerical stability in RMSProp.

    • lambda

      protected double lambda
      The L2 regularization factor, which is also the weight decay factor.

    • clipValue

      protected double clipValue
      The gradient clipping value.

    • clipNorm

      protected double clipNorm
      The gradient clipping norm.

    • t

      protected int t
      The training iterations.

  • Constructor Details

    • MultilayerPerceptron

      public MultilayerPerceptron(Layer... net)
      Constructor.
      Parameters:
      net - the input layer, hidden layers, and output layer in order.

  • Method Details

    • close

      public void close()
      Specified by:
      close in interface AutoCloseable

    • toString

      public String toString()
      Overrides:
      toString in class Object

    • setLearningRate

      public void setLearningRate(TimeFunction rate)
      Sets the learning rate.
      Parameters:
      rate - the learning rate.

    • setMomentum

      public void setMomentum(TimeFunction momentum)
      Sets the momentum factor. momentum = 0.0 means no momentum.
      Parameters:
      momentum - the momentum factor.

    • setRMSProp

      public void setRMSProp(double rho, double epsilon)
      Sets RMSProp parameters.
      Parameters:
      rho - The discounting factor for the history/coming gradient.
      epsilon - A small constant for numerical stability.

    • setWeightDecay

      public void setWeightDecay(double lambda)
      Sets the weight decay factor. After each weight update, every weight is simply "decayed" or shrunk according to w = w * (1 - 2 * eta * lambda).
      Parameters:
      lambda - the weight decay factor.

    • setClipValue

      public void setClipValue(double clipValue)
      Sets the gradient clipping value. If clip value is set, the gradient of each weight is clipped to be no higher than this value.
      Parameters:
      clipValue - the gradient clipping value.

    • setClipNorm

      public void setClipNorm(double clipNorm)
      Sets the gradient clipping norm. If clip norm is set, the gradient of each weight is individually clipped so that its norm is no higher than this value.
      Parameters:
      clipNorm - the gradient clipping norm.

    • getLearningRate

      public double getLearningRate()
      Returns the learning rate.
      Returns:
      the learning rate.

    • getMomentum

      public double getMomentum()
      Returns the momentum factor.
      Returns:
      the momentum factor.

    • getWeightDecay

      public double getWeightDecay()
      Returns the weight decay factor.
      Returns:
      the weight decay factor.

    • getClipValue

      public double getClipValue()
      Returns the gradient clipping value.
      Returns:
      the gradient clipping value.

    • getClipNorm

      public double getClipNorm()
      Returns the gradient clipping norm.
      Returns:
      the gradient clipping norm.

    • propagate

      protected void propagate(double[] x, boolean training)
      Propagates the signals through the neural network.
      Parameters:
      x - the input signal.
      training - true if this is in training pass.

    • backpropagate

      protected void backpropagate(boolean update)
      Propagates the errors back through the network.
      Parameters:
      update - the flag if update the weights directly. It should be false for (mini-)batch.

    • update

      protected void update(int m)
      Updates the weights for mini-batch training.
      Parameters:
      m - the mini-batch size.

    • setParameters

      public void setParameters(Properties params)
      Sets MLP hyperparameters such as learning rate, weight decay, momentum, RMSProp, etc.
      Parameters:
      params - the MLP hyperparameters.