org.nd4j.finitedifferences

Class TwoPointApproximation

java.lang.Object

org.nd4j.finitedifferences.TwoPointApproximation

public class TwoPointApproximation
extends Object

Simple 2 point finite difference approximation to compute the partial derivatives wrt the 2 given points based on: https://github.com/apache/commons-math/blob/master/src/main/java/org/apache/commons/math4/analysis/interpolation/BicubicInterpolator.java

Author:

Adam Gibson

Constructor Summary

Constructors

Constructor and Description
TwoPointApproximation()

Method Summary

Modifier and Type	Method and Description
static INDArray[]	adjustSchemeToBounds(INDArray x, INDArray h, int numSteps, INDArray lowerBound, INDArray upperBound) Adjust final scheme to presence of bounds Returns (in this order): adjusted hypothesis, whether to use onesided as an int mask array
static INDArray	approximateDerivative(org.nd4j.linalg.function.Function<INDArray,INDArray> f, INDArray x, INDArray relStep, INDArray f0, INDArray bounds)
static INDArray	computeAbsoluteStep(INDArray x)
static INDArray	computeAbsoluteStep(INDArray relStep, INDArray x)
static INDArray	denseDifference(org.nd4j.linalg.function.Function<INDArray,INDArray> func, INDArray x0, INDArray f0, INDArray h, INDArray oneSided)
static double	getEpsRelativeTo(INDArray data)
static INDArray[]	prepareBounds(INDArray bounds, INDArray x) Prepare the boundaries for processing

Methods inherited from class java.lang.Object

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

Constructor Detail

TwoPointApproximation

public TwoPointApproximation()

Method Detail

prepareBounds

public static INDArray[] prepareBounds(INDArray bounds,
                                       INDArray x)

Prepare the boundaries for processing

Parameters:

bounds - the bounds

x - the input in to the approximation

Returns:

the lower and upper bounds as an array of ndarrays (in that order) of the same shape as x

adjustSchemeToBounds

public static INDArray[] adjustSchemeToBounds(INDArray x,
                                              INDArray h,
                                              int numSteps,
                                              INDArray lowerBound,
                                              INDArray upperBound)

Adjust final scheme to presence of bounds Returns (in this order): adjusted hypothesis, whether to use onesided as an int mask array

Parameters:

x - the point to estimate the derivative

h - the finite difference steps

numSteps - Number of h steps in 1 direction to implement finite difference scheme.

lowerBound - Lower bounds for independent variable variable

upperBound - Upper bounds for independent variable

Returns:

computeAbsoluteStep

public static INDArray computeAbsoluteStep(INDArray x)

Parameters:

x -

Returns:

getEpsRelativeTo

public static double getEpsRelativeTo(INDArray data)

computeAbsoluteStep

public static INDArray computeAbsoluteStep(INDArray relStep,
                                           INDArray x)

Parameters:

relStep -

x -

Returns:

approximateDerivative

public static INDArray approximateDerivative(org.nd4j.linalg.function.Function<INDArray,INDArray> f,
                                             INDArray x,
                                             INDArray relStep,
                                             INDArray f0,
                                             INDArray bounds)

Parameters:

f -

x -

relStep -

f0 -

bounds -

Returns:

denseDifference

public static INDArray denseDifference(org.nd4j.linalg.function.Function<INDArray,INDArray> func,
                                       INDArray x0,
                                       INDArray f0,
                                       INDArray h,
                                       INDArray oneSided)

Parameters:

func -

x0 -

f0 -

h -

oneSided -

Returns: