org.deeplearning4j.nn

Class LogisticRegression

java.lang.Object

org.deeplearning4j.nn.LogisticRegression

All Implemented Interfaces:

Serializable

public class LogisticRegression
extends Object
implements Serializable

Logistic regression implementation with jblas.

Author:

Adam Gibson

See Also:

Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	LogisticRegression.Builder

Constructor Summary

Constructors

Constructor and Description
LogisticRegression(org.jblas.DoubleMatrix input, org.jblas.DoubleMatrix labels, int nIn, int nOut)
LogisticRegression(org.jblas.DoubleMatrix input, int nIn, int nOut)
LogisticRegression(int nIn, int nOut)

Method Summary

Methods

Modifier and Type	Method and Description
protected LogisticRegression	clone()
org.jblas.DoubleMatrix	getB()
LogisticRegressionGradient	getGradient(double lr) Gets the gradient from one training iteration
org.jblas.DoubleMatrix	getInput()
double	getL2()
org.jblas.DoubleMatrix	getLabels()
int	getnIn()
int	getnOut()
org.jblas.DoubleMatrix	getW()
boolean	isUseRegularization()
void	merge(LogisticRegression l, int batchSize) Averages the given logistic regression from a mini batch in to this one
double	negativeLogLikelihood() Objective function: minimize negative log likelihood
org.jblas.DoubleMatrix	predict(org.jblas.DoubleMatrix x) Classify input
void	setB(org.jblas.DoubleMatrix b)
void	setInput(org.jblas.DoubleMatrix input)
void	setL2(double l2)
void	setLabels(org.jblas.DoubleMatrix labels)
void	setnIn(int nIn)
void	setnOut(int nOut)
void	setUseRegularization(boolean useRegularization)
void	setW(org.jblas.DoubleMatrix w)
void	train(double lr) Train with current input and labels with the given learning rate
void	train(org.jblas.DoubleMatrix x, double lr) Train with the given input and the currently set labels
void	train(org.jblas.DoubleMatrix x, org.jblas.DoubleMatrix y, double lr) Train on the given inputs and labels.
void	trainTillConvergence(double learningRate, int numEpochs) Run conjugate gradient
void	trainTillConvergence(org.jblas.DoubleMatrix x, org.jblas.DoubleMatrix y, double learningRate, int epochs) Run conjugate gradient with the given x and y

Methods inherited from class java.lang.Object

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

Constructor Detail

LogisticRegression

public LogisticRegression(org.jblas.DoubleMatrix input,
                  org.jblas.DoubleMatrix labels,
                  int nIn,
                  int nOut)

LogisticRegression

public LogisticRegression(org.jblas.DoubleMatrix input,
                  int nIn,
                  int nOut)

LogisticRegression

public LogisticRegression(int nIn,
                  int nOut)

Method Detail

train

public void train(double lr)

Train with current input and labels with the given learning rate

Parameters:

lr - the learning rate to use

train

public void train(org.jblas.DoubleMatrix x,
         double lr)

Train with the given input and the currently set labels

Parameters:

x - the input to use

lr - the learning rate to use

trainTillConvergence

public void trainTillConvergence(org.jblas.DoubleMatrix x,
                        org.jblas.DoubleMatrix y,
                        double learningRate,
                        int epochs)

Run conjugate gradient with the given x and y

Parameters:

x - the input to use

y - the labels to use

learningRate -

epochs -

trainTillConvergence

public void trainTillConvergence(double learningRate,
                        int numEpochs)

Run conjugate gradient

Parameters:

learningRate - the learning rate to train with

numEpochs - the number of epochs

merge

public void merge(LogisticRegression l,
         int batchSize)

Averages the given logistic regression from a mini batch in to this one

Parameters:

l - the logistic regression to average in to this one

batchSize - the batch size

negativeLogLikelihood

public double negativeLogLikelihood()

Objective function: minimize negative log likelihood

Returns:

the negative log likelihood of the model

train

public void train(org.jblas.DoubleMatrix x,
         org.jblas.DoubleMatrix y,
         double lr)

Train on the given inputs and labels. This will assign the passed in values as fields to this logistic function for caching.

Parameters:

x - the inputs to train on

y - the labels to train on

lr - the learning rate

clone

protected LogisticRegression clone()

Overrides:

clone in class Object

getGradient

public LogisticRegressionGradient getGradient(double lr)

Gets the gradient from one training iteration

Parameters:

lr - the learning rate to use for training

Returns:

the gradient (bias and weight matrix)

predict

public org.jblas.DoubleMatrix predict(org.jblas.DoubleMatrix x)

Classify input

Parameters:

x - the input (can either be a matrix or vector) If it's a matrix, each row is considered an example and associated rows are classified accordingly. Each row will be the likelihood of a label given that example

Returns:

a probability distribution for each row

getnIn

public int getnIn()

setnIn

public void setnIn(int nIn)

getnOut

public int getnOut()

setnOut

public void setnOut(int nOut)

getInput

public org.jblas.DoubleMatrix getInput()

setInput

public void setInput(org.jblas.DoubleMatrix input)

getLabels

public org.jblas.DoubleMatrix getLabels()

setLabels

public void setLabels(org.jblas.DoubleMatrix labels)

getW

public org.jblas.DoubleMatrix getW()

setW

public void setW(org.jblas.DoubleMatrix w)

getB

public org.jblas.DoubleMatrix getB()

setB

public void setB(org.jblas.DoubleMatrix b)

getL2

public double getL2()

setL2

public void setL2(double l2)

isUseRegularization

public boolean isUseRegularization()

setUseRegularization

public void setUseRegularization(boolean useRegularization)