com.workfusion.vds.sdk.ml.dp.bolton.linear

Class BaseDpLinearClassifier

java.lang.Object

com.workfusion.vds.sdk.ml.python.BasePythonClassifier

com.workfusion.vds.sdk.ml.dp.bolton.linear.BaseDpLinearClassifier

All Implemented Interfaces:

Classifier, AutoCloseable

Direct Known Subclasses:

DpLogisticRegressionClassifier, DpSvmClassifier

public abstract class BaseDpLinearClassifier
extends BasePythonClassifier

The base class for Bolton Differential Privacy Linear classifiers.

Since:

10.0

Author:

Konstantin Kovalev

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	BaseDpLinearClassifier.Builder<T extends BasePythonClassifier> The type Builder.

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	BaseDpLinearClassifier(BaseDpLinearClassifier.Builder builder)

Method Summary

Modifier and Type	Method and Description
static BaseDpLinearClassifier.Builder	builder()
int	getBatchSize() Gets the number of samples in one batch.
double	getDpEpsilon() Gets the value that controls the strength of the privacy guarantees.
double	getLearningRate() Gets learning rate.
int	getMaxEpochs() Gets the number of epochs to run for training.
double	getMomentum() Gets optimization coefficient.
double	getPenaltyCoefficient() Gets the value for the penalty parameter C of the error term.
double	getRadiusConstant() Gets the value for the radius factor on which to project the weight parameters.
double	getRegLambda() Gets the value for the L2 regularization parameter.
Object	getSampleWeights() Gets sample weights (SampleWeightsType or list of sample weights for each class type).
int	getVerbose() Gets verbose level.
boolean	isGpu() Whether the gpu is used during training.
boolean	isHighMemory() Wheather high memory mode is enabled (gpu only) where training data is kept on CPU memory and only the mini-batch data is sent.

Methods inherited from class com.workfusion.vds.sdk.ml.python.BasePythonClassifier

close, fit, getModelName, getWorkingDir, init, predict

Methods inherited from class java.lang.Object

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

Constructor Detail

BaseDpLinearClassifier

protected BaseDpLinearClassifier(BaseDpLinearClassifier.Builder builder)

Method Detail

getMaxEpochs

public int getMaxEpochs()

Gets the number of epochs to run for training.

Returns:

the max epochs

getBatchSize

public int getBatchSize()

Gets the number of samples in one batch.

Returns:

the batch size

getSampleWeights

public Object getSampleWeights()

Gets sample weights (SampleWeightsType or list of sample weights for each class type). If null then all classes are supposed to have weight one.

Returns:

the sample weights

getVerbose

public int getVerbose()

Gets verbose level. It set ups approximately how many informational messages to generate during training.

Returns:

the verbose level.

isGpu

public boolean isGpu()

Whether the gpu is used during training.

Returns:

the gpu flag

getDpEpsilon

public double getDpEpsilon()

Gets the value that controls the strength of the privacy guarantees.

Returns:

the dp epsilon

isHighMemory

public boolean isHighMemory()

Wheather high memory mode is enabled (gpu only) where training data is kept on CPU memory and only the mini-batch data is sent.

Returns:

high memory flag

getMomentum

public double getMomentum()

Gets optimization coefficient.

Returns:

the momentum

getLearningRate

public double getLearningRate()

Gets learning rate.

Returns:

the learning rate

getRadiusConstant

public double getRadiusConstant()

Gets the value for the radius factor on which to project the weight parameters.

Returns:

the radius constant

getRegLambda

public double getRegLambda()

Gets the value for the L2 regularization parameter.

Returns:

the reg lambda

getPenaltyCoefficient

public double getPenaltyCoefficient()

Gets the value for the penalty parameter C of the error term.

Returns:

the penalty coefficient

builder

public static BaseDpLinearClassifier.Builder builder()