Class Ftrl

java.lang.Object

org.tensorflow.framework.optimizers.Optimizer

org.tensorflow.framework.optimizers.Ftrl

public class Ftrl
extends Optimizer

Optimizer that implements the FTRL algorithm.

This version has support for both online L2 (the L2 penalty given in the paper below) and shrinkage-type L2 (which is the addition of an L2 penalty to the loss function).

See Also:

• McMahan, et al., 2013, Algorithm 1

Nested Class Summary

Nested classes/interfaces inherited from class Optimizer

Optimizer.GradAndVar<T>, Optimizer.Options

Field Summary

Fields

Modifier and Type	Field	Description
static final String	ACCUMULATOR
static final float	INITIAL_ACCUMULATOR_VALUE_DEFAULT
static final float	L1STRENGTH_DEFAULT
static final float	L2_SHRINKAGE_REGULARIZATION_STRENGTH_DEFAULT
static final float	L2STRENGTH_DEFAULT
static final float	LEARNING_RATE_DEFAULT
static final float	LEARNING_RATE_POWER_DEFAULT
static final String	LINEAR_ACCUMULATOR

Fields inherited from class Optimizer

globals, graph, tf, VARIABLE_V2

Constructor Summary

Constructors

Constructor	Description
Ftrl(Graph graph)	Creates a Ftrl Optimizer
Ftrl(Graph graph, float learningRate)	Creates a Ftrl Optimizer
Ftrl(Graph graph, float learningRate, float learningRatePower, float initialAccumulatorValue, float l1Strength, float l2Strength, float l2ShrinkageRegularizationStrength)	Creates a Ftrl Optimizer
Ftrl(Graph graph, String name)	Creates a Ftrl Optimizer
Ftrl(Graph graph, String name, float learningRate)	Creates a Ftrl Optimizer
Ftrl(Graph graph, String name, float learningRate, float learningRatePower, float initialAccumulatorValue, float l1Strength, float l2Strength, float l2ShrinkageRegularizationStrength)	Creates a Ftrl Optimizer

Method Summary

Modifier and Type	Method	Description
protected <T extends TType> Op	applyDense(Ops deps, Output<T> gradient, Output<T> variable)	Generates the gradient update operations for the specific variable and gradient.
protected void	createSlots(List<Output<? extends TType>> variables)	Performs a No-op slot creation method.
String	getOptimizerName()	Get the Name of the optimizer.

Methods inherited from class Optimizer

applyGradients, computeGradients, createName, createSlot, finish, getSlot, getTF, minimize, minimize, prepare

Methods inherited from class Object

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

Field Details

ACCUMULATOR

public static final String ACCUMULATOR

See Also:

• Constant Field Values

LINEAR_ACCUMULATOR

public static final String LINEAR_ACCUMULATOR

See Also:

• Constant Field Values

LEARNING_RATE_DEFAULT

public static final float LEARNING_RATE_DEFAULT

See Also:

• Constant Field Values

LEARNING_RATE_POWER_DEFAULT

public static final float LEARNING_RATE_POWER_DEFAULT

See Also:

• Constant Field Values

INITIAL_ACCUMULATOR_VALUE_DEFAULT

public static final float INITIAL_ACCUMULATOR_VALUE_DEFAULT

See Also:

• Constant Field Values

L1STRENGTH_DEFAULT

public static final float L1STRENGTH_DEFAULT

See Also:

• Constant Field Values

L2STRENGTH_DEFAULT

public static final float L2STRENGTH_DEFAULT

See Also:

• Constant Field Values

L2_SHRINKAGE_REGULARIZATION_STRENGTH_DEFAULT

public static final float L2_SHRINKAGE_REGULARIZATION_STRENGTH_DEFAULT

See Also:

• Constant Field Values

Constructor Details

Ftrl

public Ftrl(Graph graph)

Creates a Ftrl Optimizer

Parameters:

graph - the TensorFlow Graph

Ftrl

public Ftrl(Graph graph,
 String name)

Creates a Ftrl Optimizer

Parameters:

graph - the TensorFlow Graph

name - the name of this Optimizer

Ftrl

public Ftrl(Graph graph,
 float learningRate)

Creates a Ftrl Optimizer

Parameters:

graph - the TensorFlow Graph

learningRate - the learning rate

Ftrl

public Ftrl(Graph graph,
 String name,
 float learningRate)

Creates a Ftrl Optimizer

Parameters:

graph - the TensorFlow Graph

name - the name of this Optimizer

learningRate - the learning rate

Ftrl

public Ftrl(Graph graph,
 float learningRate,
 float learningRatePower,
 float initialAccumulatorValue,
 float l1Strength,
 float l2Strength,
 float l2ShrinkageRegularizationStrength)

Creates a Ftrl Optimizer

Parameters:

graph - the TensorFlow Graph

learningRate - the learning rate

learningRatePower - Controls how the learning rate decreases during training. Use zero for a fixed learning rate.

initialAccumulatorValue - The starting value for accumulators. Only zero or positive values are allowed.

l1Strength - the L1 Regularization strength, must be greater than or equal to zero.

l2Strength - the L2 Regularization strength, must be greater than or equal to zero.

l2ShrinkageRegularizationStrength - This differs from L2 above in that the L2 above is a stabilization penalty, whereas this L2 shrinkage is a magnitude penalty. must be greater than or equal to zero.

Throws:

IllegalArgumentException - if the initialAccumulatorValue, l1RegularizationStrength, l2RegularizationStrength, or l2ShrinkageRegularizationStrength are less than 0.0, or learningRatePower is greater than 0.0.

Ftrl

public Ftrl(Graph graph,
 String name,
 float learningRate,
 float learningRatePower,
 float initialAccumulatorValue,
 float l1Strength,
 float l2Strength,
 float l2ShrinkageRegularizationStrength)

Creates a Ftrl Optimizer

Parameters:

graph - the TensorFlow Graph

name - the name of this Optimizer

learningRate - the learning rate

learningRatePower - Controls how the learning rate decreases during training. Use zero for a fixed learning rate.

initialAccumulatorValue - The starting value for accumulators. Only zero or positive values are allowed.

l1Strength - the L1 Regularization strength, must be greater than or equal to zero.

l2Strength - the L2 Regularization strength, must be greater than or equal to zero.

l2ShrinkageRegularizationStrength - This differs from L2 above in that the L2 above is a stabilization penalty, whereas this L2 shrinkage is a magnitude penalty. must be greater than or equal to zero.

Throws:

IllegalArgumentException - if the initialAccumulatorValue, l1RegularizationStrength, l2RegularizationStrength, or l2ShrinkageRegularizationStrength are less than 0.0, or learningRatePower is greater than 0.0.

Method Details

createSlots

protected void createSlots(List<Output<? extends TType>> variables)

Performs a No-op slot creation method.

Overrides:

createSlots in class Optimizer

Parameters:

variables - The variables to create slots for.

applyDense

protected <T extends TType> Op applyDense(Ops deps,
 Output<T> gradient,
 Output<T> variable)

Generates the gradient update operations for the specific variable and gradient.

Specified by:

applyDense in class Optimizer

Type Parameters:

T - The type of the variable.

Parameters:

gradient - The gradient to use.

variable - The variable to update.

Returns:

An operand which applies the desired optimizer update to the variable.

getOptimizerName

public String getOptimizerName()

Get the Name of the optimizer.

Specified by:

getOptimizerName in class Optimizer

Returns:

The optimizer name.