BackPropagationNet (Java Statistical Analysis Tool 0.0.8 API)

java.lang.Object
- jsat.classifiers.neuralnetwork.BackPropagationNet

All Implemented Interfaces:

Serializable, Cloneable, Classifier, Parameterized, Regressor
```
public class BackPropagationNet
extends Object
implements Classifier, Regressor, Parameterized
```
An implementation of a Feed Forward Neural Network (NN) trained by Back Propagation. NNs are powerful classifiers and regressors, but can suffer from slow training time and overfitting.

NOTE: This class should generally not be used any more. The DReDNetSimple provides an easier to use class for most cases that will likely converge faster.

Author:

Edward Raff

See Also:

Serialized Form

Nested Class Summary

Nested Classes
Modifier and Type	Class and Description
`static class`	`BackPropagationNet.ActivationFunction` The neural network needs an activation function for the neurons that is used to predict from inputs and train the network by propagating the errors back through the network.
`static class`	`BackPropagationNet.WeightInitialization` Different methods of initializing the weight values before training

Field Summary

Fields
Modifier and Type	Field and Description
`static BackPropagationNet.ActivationFunction`	`logitActiv` The logit activation function.
`static BackPropagationNet.ActivationFunction`	`softsignActiv` The softsign activation function.
`static BackPropagationNet.ActivationFunction`	`tanhActiv` The tanh activation function.

Constructor Summary

Constructors
Modifier	Constructor and Description
	`BackPropagationNet()` Creates a new back propagation network with one hidden layer of 1024 neurons.
`protected`	`BackPropagationNet(BackPropagationNet toClone)` Copy constructor
	`BackPropagationNet(int... npl)` Creates a new back propagation network.

Method Summary

All Methods Instance Methods Concrete Methods
Modifier and Type	Method and Description
`CategoricalResults`	`classify(DataPoint data)` Performs classification on the given data point.
`BackPropagationNet`	`clone()`
`BackPropagationNet.ActivationFunction`	`getActivationFunction()` Returns the activation function used for training the network
`int`	`getBatchSize()` Returns the training batch size
`int`	`getEpochs()` Returns the number of epochs of training epochs for learning
`double`	`getInitialLearningRate()` Returns the learning rate used
`DecayRate`	`getLearningRateDecay()` Returns the decay rate used to adjust the learning rate after each epoch
`double`	`getMomentum()` Returns the momentum in use
`Parameter`	`getParameter(String paramName)` Returns the parameter with the given name.
`List<Parameter>`	`getParameters()` Returns the list of parameters that can be altered for this learner.
`double`	`getWeightDecay()` Returns the weight decay used for each update
`BackPropagationNet.WeightInitialization`	`getWeightInitialization()` Returns the method of weight initialization used
`double`	`regress(DataPoint data)`
`void`	`setActivationFunction(BackPropagationNet.ActivationFunction f)` Sets the activation function used for the network
`void`	`setBatchSize(int batchSize)` Sets the batch size use to estimate the gradient of the error for training
`void`	`setEpochs(int epochs)` Sets the number of epochs of training used.
`void`	`setInitialLearningRate(double initialLearningRate)` Sets the initial learning rate used for the first epoch
`void`	`setLearningRateDecay(DecayRate learningRateDecay)` Sets the decay rate used to adjust the learning rate after each epoch
`void`	`setMomentum(double momentum)` Sets the non negative momentum used in training.
`void`	`setWeightDecay(double weightDecay)` Sets the weight decay used for each update.
`void`	`setWeightInitialization(BackPropagationNet.WeightInitialization weightInitialization)` Sets how the weights are initialized before training starts
`boolean`	`supportsWeightedData()` Indicates whether the model knows how to train using weighted data points.
`void`	`train(RegressionDataSet dataSet)`
`void`	`train(RegressionDataSet dataSet, ExecutorService threadPool)`
`void`	`trainC(ClassificationDataSet dataSet)` Trains the classifier and constructs a model for classification using the given data set.
`void`	`trainC(ClassificationDataSet dataSet, ExecutorService threadPool)` Trains the classifier and constructs a model for classification using the given data set.

Methods inherited from class java.lang.Object
equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

- Field Detail
  - logitActiv
```
public static final BackPropagationNet.ActivationFunction logitActiv
```
    The logit activation function. This function goes from [0, 1]. It has more difficultly learning than symmetric activation functions, often requiring considerably more layers and neurons than other activation functions.
  - tanhActiv
```
public static final BackPropagationNet.ActivationFunction tanhActiv
```
    The tanh activation function. This function is symmetric in the range of [-1, 1]. It works well for many problems in general.
  - softsignActiv
```
public static final BackPropagationNet.ActivationFunction softsignActiv
```
    The softsign activation function. This function is symmetric in the range of [-1, 1]. It works well for classification problems, and is very fast to compute. It sometimes requires more neurons to learn more complicated functions / boundaries. It sometimes has reduced performance in regression
- Constructor Detail
  - BackPropagationNet
```
public BackPropagationNet()
```
    Creates a new back propagation network with one hidden layer of 1024 neurons.
    
    Parameters:
    
    npl - the array of hidden layer information. The length indicates how many hidden layers, and the value of each index indicates how many neurons to place in each hidden layer
  - BackPropagationNet
```
public BackPropagationNet(int... npl)
```
    Creates a new back propagation network.
    
    Parameters:
    
    npl - the array of hidden layer information. The length indicates how many hidden layers, and the value of each index indicates how many neurons to place in each hidden layer
  - BackPropagationNet
```
protected BackPropagationNet(BackPropagationNet toClone)
```
    Copy constructor
    
    Parameters:
    
    toClone - the one to copy
- Method Detail
  - setMomentum
```
public void setMomentum(double momentum)
```
    Sets the non negative momentum used in training.
    
    Parameters:
    
    momentum - the momentum to apply to training
  - getMomentum
```
public double getMomentum()
```
    Returns the momentum in use
    
    Returns:
    
    the momentum
  - setInitialLearningRate
```
public void setInitialLearningRate(double initialLearningRate)
```
    Sets the initial learning rate used for the first epoch
    
    Parameters:
    
    initialLearningRate - the positive learning rate to use
  - getInitialLearningRate
```
public double getInitialLearningRate()
```
    Returns the learning rate used
    
    Returns:
    
    the learning rate used
  - setLearningRateDecay
```
public void setLearningRateDecay(DecayRate learningRateDecay)
```
    Sets the decay rate used to adjust the learning rate after each epoch
    
    Parameters:
    
    learningRateDecay - the decay for the learning rate
  - getLearningRateDecay
```
public DecayRate getLearningRateDecay()
```
    Returns the decay rate used to adjust the learning rate after each epoch
    
    Returns:
    
    the decay rate used for learning
  - setEpochs
```
public void setEpochs(int epochs)
```
    Sets the number of epochs of training used. Each epoch goes through the whole data set once.
    
    Parameters:
    
    epochs - the number of training epochs
  - getEpochs
```
public int getEpochs()
```
    Returns the number of epochs of training epochs for learning
    
    Returns:
    
    the number of training epochs
  - setWeightDecay
```
public void setWeightDecay(double weightDecay)
```
    Sets the weight decay used for each update. The weight decay must be in the range [0, 1). Weight decay values must often be very small, often 1e-8 or less.
    
    Parameters:
    
    weightDecay - the weight decay to apply when training
  - getWeightDecay
```
public double getWeightDecay()
```
    Returns the weight decay used for each update
    
    Returns:
    
    the weight decay used.
  - setWeightInitialization
```
public void setWeightInitialization(BackPropagationNet.WeightInitialization weightInitialization)
```
    Sets how the weights are initialized before training starts
    
    Parameters:
    
    weightInitialization - the method of weight initialization
  - getWeightInitialization
```
public BackPropagationNet.WeightInitialization getWeightInitialization()
```
    Returns the method of weight initialization used
    
    Returns:
    
    the method of weight initialization used
  - setBatchSize
```
public void setBatchSize(int batchSize)
```
    Sets the batch size use to estimate the gradient of the error for training
    
    Parameters:
    
    batchSize - the number of training instances to use on each update
  - getBatchSize
```
public int getBatchSize()
```
    Returns the training batch size
    
    Returns:
    
    the batch size used for training
  - setActivationFunction
```
public void setActivationFunction(BackPropagationNet.ActivationFunction f)
```
    Sets the activation function used for the network
    
    Parameters:
    
    f - the activation function to use
  - getActivationFunction
```
public BackPropagationNet.ActivationFunction getActivationFunction()
```
    Returns the activation function used for training the network
    
    Returns:
    
    the activation function in use
  - classify
```
public CategoricalResults classify(DataPoint data)
```
    Description copied from interface: Classifier
    
    Performs classification on the given data point.
    
    Specified by:
    
    classify in interface Classifier
    
    Parameters:
    
    data - the data point to classify
    
    Returns:
    
    the results of the classification.
  - regress
```
public double regress(DataPoint data)
```
    Specified by:
    
    regress in interface Regressor
  - trainC
```
public void trainC(ClassificationDataSet dataSet,
                   ExecutorService threadPool)
```
    Description copied from interface: Classifier
    
    Trains the classifier and constructs a model for classification using the given data set. If the training method knows how, it will used the threadPool to conduct training in parallel. This method will block until the training has completed.
    
    Specified by:
    
    trainC in interface Classifier
    
    Parameters:
    
    dataSet - the data set to train on
    
    threadPool - the source of threads to use.
  - trainC
```
public void trainC(ClassificationDataSet dataSet)
```
    Description copied from interface: Classifier
    
    Trains the classifier and constructs a model for classification using the given data set.
    
    Specified by:
    
    trainC in interface Classifier
    
    Parameters:
    
    dataSet - the data set to train on
  - train
```
public void train(RegressionDataSet dataSet,
                  ExecutorService threadPool)
```
    Specified by:
    
    train in interface Regressor
  - train
```
public void train(RegressionDataSet dataSet)
```
    Specified by:
    
    train in interface Regressor
  - supportsWeightedData
```
public boolean supportsWeightedData()
```
    Description copied from interface: Classifier
    
    Indicates whether the model knows how to train using weighted data points. If it does, the model will train assuming the weights. The values returned by this method may change depending on the parameters set for the model.
    
    Specified by:
    
    supportsWeightedData in interface Classifier
    
    Specified by:
    
    supportsWeightedData in interface Regressor
    
    Returns:
    
    true if the model supports weighted data, false otherwise
  - clone
```
public BackPropagationNet clone()
```
    Specified by:
    
    clone in interface Classifier
    
    Specified by:
    
    clone in interface Regressor
    
    Overrides:
    
    clone in class Object
  - getParameters
```
public List<Parameter> getParameters()
```
    Description copied from interface: Parameterized
    
    Returns the list of parameters that can be altered for this learner.
    
    Specified by:
    
    getParameters in interface Parameterized
    
    Returns:
    
    the list of parameters that can be altered for this learner.
  - getParameter
```
public Parameter getParameter(String paramName)
```
    Description copied from interface: Parameterized
    
    Returns the parameter with the given name. Two different strings may map to a single Parameter object. An ASCII only string, and a Unicode style string.
    
    Specified by:
    
    getParameter in interface Parameterized
    
    Parameters:
    
    paramName - the name of the parameter to obtain
    
    Returns:
    
    the Parameter in question, or null if no such named Parameter exists.

Class BackPropagationNet

Nested Class Summary

Field Summary

Constructor Summary

Method Summary

Methods inherited from class java.lang.Object

Field Detail

logitActiv

tanhActiv

softsignActiv

Constructor Detail

BackPropagationNet

BackPropagationNet

BackPropagationNet

Method Detail

setMomentum

getMomentum

setInitialLearningRate

getInitialLearningRate

setLearningRateDecay

getLearningRateDecay

setEpochs

getEpochs

setWeightDecay

getWeightDecay

setWeightInitialization

getWeightInitialization

setBatchSize

getBatchSize

setActivationFunction

getActivationFunction

classify

regress

trainC

trainC

train

train

supportsWeightedData

clone

getParameters

getParameter