mikera.vectorz

Class Op

java.lang.Object

mikera.vectorz.Op

All Implemented Interfaces:

ITransform, IOp

Direct Known Subclasses:

ABoundedOp, ALinearOp, ComposedOp, GaussianNoise, InverseOp

public abstract class Op
extends Object
implements IOp, ITransform

Abstract class for representing a unary operation

Author:

Mike

Field Summary

Fields

Modifier and Type	Field and Description
static Op	LINEAR
static Op	LOGISTIC
static Op	RECTIFIER
static Op	SOFTPLUS
static Op	STOCHASTIC_BINARY
static Op	STOCHASTIC_LOGISTIC
static Op	TANH

Constructor Summary

Constructors

Constructor and Description
Op()

Method Summary

Methods

Modifier and Type	Method and Description
abstract double	apply(double x) Applies the operator to a single value, returning the result
double	applyInverse(double y) Applies the inverse of this Op.
void	applyTo(ArrayVector v)
void	applyTo(AScalar s)
void	applyTo(AVector v)
void	applyTo(double[] data)
void	applyTo(double[] data, int start, int length)
abstract double	averageValue()
Op	compose(Op op)
static Op	compose(Op op1, Op op2)
void	constrainValues(double[] src, double[] dest, int offset, int length) Creates a copy of the values of src in dest, constraining them to be within the valid range of output values from this Op
double	derivative(double x) Returns the derivative of this Op for a given input value x i.e.
double	derivativeForOutput(double y) Returns the derivative of this Op for a given output value y i.e.
Op	getInverse()
ATransform	getTransform(int dims) Converts an operator into a corresponding transform that applies the operator to all elements of its input
boolean	hasDerivative()
boolean	hasInverse()
int	inputDimensions() Returns the number of dimensions required for input vectors
boolean	isBounded()
boolean	isStochastic() Returns true if the operator is stochastic, i.e returns random values for at least some inputs
double	maxValue()
double	minValue()
int	outputDimensions() Returns the number of dimensions required for output vectors
void	transform(AVector source, AVector dest) Transforms the source vector, storing the result in the given destination vector
boolean	validateOutput(double[] output) Validates whether all values in a double[] are within the possible output range for this Op

Methods inherited from class java.lang.Object

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

Field Detail

STOCHASTIC_BINARY

public static final Op STOCHASTIC_BINARY

LINEAR

public static final Op LINEAR

LOGISTIC

public static final Op LOGISTIC

RECTIFIER

public static final Op RECTIFIER

STOCHASTIC_LOGISTIC

public static final Op STOCHASTIC_LOGISTIC

TANH

public static final Op TANH

SOFTPLUS

public static final Op SOFTPLUS

Constructor Detail

Op

public Op()

Method Detail

apply

public abstract double apply(double x)

Description copied from interface: IOp

Applies the operator to a single value, returning the result

Specified by:

apply in interface IOp

Returns:

applyInverse

public double applyInverse(double y)

Applies the inverse of this Op. Throws an error if the inverse function does not exist. Returns Double.NaN if no inverse exists for the specific value of y.

Parameters:

y -

Returns:

applyTo

public void applyTo(AVector v)

Specified by:

applyTo in interface IOp

applyTo

public void applyTo(AScalar s)

applyTo

public void applyTo(ArrayVector v)

applyTo

public void applyTo(double[] data,
           int start,
           int length)

Specified by:

applyTo in interface IOp

applyTo

public void applyTo(double[] data)

getTransform

public ATransform getTransform(int dims)

Description copied from interface: IOp

Converts an operator into a corresponding transform that applies the operator to all elements of its input

Specified by:

getTransform in interface IOp

Returns:

inputDimensions

public int inputDimensions()

Description copied from interface: ITransform

Returns the number of dimensions required for input vectors

Specified by:

inputDimensions in interface ITransform

Returns:

outputDimensions

public int outputDimensions()

Description copied from interface: ITransform

Returns the number of dimensions required for output vectors

Specified by:

outputDimensions in interface ITransform

Returns:

transform

public void transform(AVector source,
             AVector dest)

Description copied from interface: ITransform

Transforms the source vector, storing the result in the given destination vector

Specified by:

transform in interface ITransform

getInverse

public Op getInverse()

Specified by:

getInverse in interface IOp

hasDerivative

public boolean hasDerivative()

hasInverse

public boolean hasInverse()

derivativeForOutput

public double derivativeForOutput(double y)

Returns the derivative of this Op for a given output value y i.e. f'(g(y)) where f is the operator, g is the inverse of f

Parameters:

y -

Returns:

derivative

public double derivative(double x)

Returns the derivative of this Op for a given input value x i.e. f'(x) where f is the operator

Parameters:

y -

Returns:

isStochastic

public boolean isStochastic()

Returns true if the operator is stochastic, i.e returns random values for at least some inputs

Returns:

minValue

public double minValue()

averageValue

public abstract double averageValue()

maxValue

public double maxValue()

validateOutput

public boolean validateOutput(double[] output)

Validates whether all values in a double[] are within the possible output range for this Op

Parameters:

output -

Returns:

constrainValues

public void constrainValues(double[] src,
                   double[] dest,
                   int offset,
                   int length)

Creates a copy of the values of src in dest, constraining them to be within the valid range of output values from this Op

Parameters:

src -

dest -

offset -

length -

isBounded

public boolean isBounded()

compose

public static Op compose(Op op1,
         Op op2)

compose

public Op compose(Op op)