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

Class BaseDpLinearClassifier.Builder<T extends BasePythonClassifier>

java.lang.Object

com.workfusion.vds.sdk.ml.python.BasePythonClassifier.Builder<BaseDpLinearClassifier.Builder<T>,T>

com.workfusion.vds.sdk.ml.dp.bolton.linear.BaseDpLinearClassifier.Builder<T>

Direct Known Subclasses:

DpLogisticRegressionClassifier.Builder, DpSvmClassifier.Builder

Enclosing class:

BaseDpLinearClassifier

public abstract static class BaseDpLinearClassifier.Builder<T extends BasePythonClassifier>
extends BasePythonClassifier.Builder<BaseDpLinearClassifier.Builder<T>,T>

The type Builder.

Field Summary

Fields inherited from class com.workfusion.vds.sdk.ml.python.BasePythonClassifier.Builder

modelName, workingDir

Constructor Summary

Constructors

Constructor and Description
Builder() Instantiates a new Builder.

Method Summary

Modifier and Type	Method and Description
BaseDpLinearClassifier.Builder<T>	batchSize(int batchSize) Sets the number of samples in one batch.
BaseDpLinearClassifier.Builder<T>	dpEpsilon(double dpEpsilon) Sets the DP epsilon.
BaseDpLinearClassifier.Builder<T>	gpu(boolean gpu) Sets whether gpu is used.
BaseDpLinearClassifier.Builder<T>	highMemory(boolean highMemory) Sets whether high memory mode is used.
BaseDpLinearClassifier.Builder<T>	learningRate(double learningRate) Sets the learning rate.
BaseDpLinearClassifier.Builder<T>	maxEpochs(int maxEpochs) Sets the number of epochs.
BaseDpLinearClassifier.Builder<T>	modelName(String name) Set a model name.
BaseDpLinearClassifier.Builder<T>	momentum(double momentum) Sets the momentum.
BaseDpLinearClassifier.Builder<T>	penaltyCoefficient(double penaltyCoefficient) Sets the penaltyCoefficient value.
BaseDpLinearClassifier.Builder<T>	radiusConstant(double radiusConstant) Sets the radiusConstant value.
BaseDpLinearClassifier.Builder<T>	regLambda(double regLambda) Sets the regLambda value.
BaseDpLinearClassifier.Builder<T>	sampleWeights(List<Double> sampleWeights) Sets list of sample weights for each class type.
BaseDpLinearClassifier.Builder<T>	sampleWeights(SampleWeightsType sampleWeights) Sets sample weights.
BaseDpLinearClassifier.Builder<T>	verbose(int verbose) Sets verbose level.

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

build, createDefaultClassifierBridge, workingDir

Methods inherited from class java.lang.Object

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

Constructor Detail

Builder

public Builder()

Instantiates a new Builder.

Method Detail

maxEpochs

public BaseDpLinearClassifier.Builder<T> maxEpochs(int maxEpochs)

Sets the number of epochs.

Parameters:

maxEpochs - the number of epochs to run for training (default: 1000)

Returns:

the builder

batchSize

public BaseDpLinearClassifier.Builder<T> batchSize(int batchSize)

Sets the number of samples in one batch.

Parameters:

batchSize - the number of samples in one mini-batch of SGD (default: 0 (mini-batch is the entire data set))

Returns:

the builder

sampleWeights

public BaseDpLinearClassifier.Builder<T> sampleWeights(SampleWeightsType sampleWeights)

Sets sample weights. If null then all classes are supposed to have weight one.

Parameters:

sampleWeights - the sample weights (SampleWeightsType) (default: null)

Returns:

the builder

sampleWeights

public BaseDpLinearClassifier.Builder<T> sampleWeights(List<Double> sampleWeights)

Sets list of sample weights for each class type.

Parameters:

sampleWeights - the sample weights

Returns:

the builder

dpEpsilon

public BaseDpLinearClassifier.Builder<T> dpEpsilon(double dpEpsilon)

Sets the DP epsilon.

Parameters:

dpEpsilon - the decimal value that controls the strength of the privacy guarantees (default: 2)

Returns:

the builder

verbose

public BaseDpLinearClassifier.Builder<T> verbose(int verbose)

Sets verbose level.

Parameters:

verbose - approximately how many informational messages to generate during training. For example, a value of 2 means that informational messages will be logged at the halfway point and at the end of training (default: 1).

Returns:

the builder

gpu

public BaseDpLinearClassifier.Builder<T> gpu(boolean gpu)

Sets whether gpu is used.

Parameters:

gpu - flag to use gpu (default: false).

Returns:

the builder

highMemory

public BaseDpLinearClassifier.Builder<T> highMemory(boolean highMemory)

Sets whether high memory mode is used.

Parameters:

highMemory - enable high memory mode where training data is kept on CPU memory and only the mini-batch data is sent to GPU memory for training (default: false).

Returns:

the builder

momentum

public BaseDpLinearClassifier.Builder<T> momentum(double momentum)

Sets the momentum.

Parameters:

momentum - the optimization coefficient (default: 0.0)

Returns:

the builder

learningRate

public BaseDpLinearClassifier.Builder<T> learningRate(double learningRate)

Sets the learning rate.

Parameters:

learningRate - the learning rate (default: 1.0)

Returns:

the builder

radiusConstant

public BaseDpLinearClassifier.Builder<T> radiusConstant(double radiusConstant)

Sets the radiusConstant value.

Parameters:

radiusConstant - the value for the radius factor on which to project the weight parameters (default: 1.0).

Returns:

the builder

regLambda

public BaseDpLinearClassifier.Builder<T> regLambda(double regLambda)

Sets the regLambda value.

Parameters:

regLambda - the value for the L2 regularization parameter. Cannot be 0 in order for the L2 penalizer to make the loss strongly convex (default: 1.0).

Returns:

the builder

penaltyCoefficient

public BaseDpLinearClassifier.Builder<T> penaltyCoefficient(double penaltyCoefficient)

Sets the penaltyCoefficient value.

Parameters:

penaltyCoefficient - the value for the penalty parameter C of the error term (default: 1.0).

Returns:

the builder

modelName

public BaseDpLinearClassifier.Builder<T> modelName(String name)

Description copied from class: BasePythonClassifier.Builder

Set a model name.

Overrides:

modelName in class BasePythonClassifier.Builder<BaseDpLinearClassifier.Builder<T extends BasePythonClassifier>,T extends BasePythonClassifier>

Parameters:

name - the model name (String)

Returns:

the builder