org.nd4j.linalg.api.ops

Class BaseGradientOp

java.lang.Object

org.nd4j.autodiff.functions.DifferentialFunction

org.nd4j.linalg.api.ops.BaseOp

org.nd4j.linalg.api.ops.BaseTransformOp

org.nd4j.linalg.api.ops.BaseGradientOp

All Implemented Interfaces:

GradientOp, Op, TransformOp

Direct Known Subclasses:

GradientBackwardsMarker, LogSigmoidDerivative, LogSoftMaxDerivative, SigmoidDerivative, SoftMaxDerivative, TanhDerivative

public abstract class BaseGradientOp
extends BaseTransformOp
implements GradientOp

A gradient op always makes the following assumptions: there is always a y (beacuse of backpropagating or using the chain rule) and that it is special exec (for now) This op opType sis meant to be used to build derivative operations.

Author:

Adam Gibson

Nested Class Summary

Nested classes/interfaces inherited from interface org.nd4j.linalg.api.ops.Op

Op.Type

Field Summary

Fields inherited from class org.nd4j.linalg.api.ops.BaseOp

extraArgs, extraArgz, n, numProcessed, passThrough, x, xVertexId, y, yVertexId, z, zVertexId

Fields inherited from class org.nd4j.autodiff.functions.DifferentialFunction

dimensions, inPlace, sameDiff, scalarValue

Constructor Summary

Constructors

Constructor and Description
BaseGradientOp()
BaseGradientOp(INDArray x)
BaseGradientOp(INDArray x, INDArray z)
BaseGradientOp(INDArray x, INDArray y, INDArray z, long n)
BaseGradientOp(INDArray x, INDArray z, long n)
BaseGradientOp(SameDiff sameDiff, SDVariable i_v1, SDVariable i_v2)
BaseGradientOp(SameDiff sameDiff, SDVariable i_v1, SDVariable i_v2, boolean inPlace)

Method Summary

Modifier and Type	Method and Description
boolean	isExecSpecial() Whether the executioner needs to do a special call or not
boolean	isPassThrough() Returns whether the op should be executed or not (through the executioner)
INDArray	wrt() The array to the gradient with respect to

Methods inherited from class org.nd4j.linalg.api.ops.BaseTransformOp

calculateOutputShape, opType, z

Methods inherited from class org.nd4j.linalg.api.ops.BaseOp

equals, exec, exec, extraArgs, extraArgsBuff, extraArgsDataBuff, getOpType, hashCode, init, initFromOnnx, initFromTensorFlow, n, numProcessed, outputVariables, setN, setX, setY, setZ, toCustomOp, toString, x, y

Methods inherited from class org.nd4j.autodiff.functions.DifferentialFunction

arg, args, asProperties, attributeAdaptersForFunction, configFieldName, diff, doDiff, dup, f, getValue, hasPlaceHolderInputs, isConfigProperties, larg, mappingsForFunction, onnxName, onnxNames, opName, opNum, outputVariables, propertiesForFunction, rarg, resolvePropertiesFromSameDiffBeforeExecution, setInstanceId, setValueFor, tensorflowName, tensorflowNames

Methods inherited from class java.lang.Object

clone, finalize, getClass, notify, notifyAll, wait, wait, wait

Methods inherited from interface org.nd4j.linalg.api.ops.Op

exec, exec, extraArgs, extraArgsBuff, extraArgsDataBuff, init, n, numProcessed, opName, opNum, setExtraArgs, setN, setX, setY, setZ, toCustomOp, x, y, z

Constructor Detail

BaseGradientOp

public BaseGradientOp(SameDiff sameDiff,
                      SDVariable i_v1,
                      SDVariable i_v2)

BaseGradientOp

public BaseGradientOp(SameDiff sameDiff,
                      SDVariable i_v1,
                      SDVariable i_v2,
                      boolean inPlace)

BaseGradientOp

public BaseGradientOp(INDArray x,
                      INDArray z)

BaseGradientOp

public BaseGradientOp()

BaseGradientOp

public BaseGradientOp(INDArray x,
                      INDArray z,
                      long n)

BaseGradientOp

public BaseGradientOp(INDArray x,
                      INDArray y,
                      INDArray z,
                      long n)

BaseGradientOp

public BaseGradientOp(INDArray x)

Method Detail

wrt

public INDArray wrt()

The array to the gradient with respect to

Specified by:

wrt in interface GradientOp

Returns:

isExecSpecial

public boolean isExecSpecial()

Description copied from interface: Op

Whether the executioner needs to do a special call or not

Specified by:

isExecSpecial in interface Op

Overrides:

isExecSpecial in class BaseOp

Returns:

true if the executioner needs to do a special call or not false otherwise

isPassThrough

public boolean isPassThrough()

Description copied from interface: Op

Returns whether the op should be executed or not (through the executioner)

Specified by:

isPassThrough in interface Op

Overrides:

isPassThrough in class BaseOp

Returns:

true if the op is pass through false otherwise