org.nd4j.linalg.api.ops.factory

Interface OpFactory

All Known Implementing Classes:

DefaultOpFactory

public interface OpFactory

Op factory

Author:

Adam Gibson

Method Summary

Modifier and Type	Method and Description
Accumulation	createAccum(String name, INDArray x) Accumulation operation
Accumulation	createAccum(String name, INDArray x, INDArray y)
Accumulation	createAccum(String name, INDArray x, INDArray y, INDArray z) Accumulation operation
BroadcastOp	createBroadcastOp(String name, INDArray x, INDArray y, INDArray z, int... dimension) Create a vector operation
BroadcastOp	createBroadcastOp(String name, INDArray x, INDArray z, int... dimension) Create a vector operation
IndexAccumulation	createIndexAccum(String name, INDArray x) Index accumulation operation
IndexAccumulation	createIndexAccum(String name, INDArray x, INDArray y) Index accumulation operation
LossFunction	createLossFunction(String name, INDArray x, INDArray y) Create a loss function with the given inputs and outputs
TransformOp	createTransform(String name, INDArray x)
TransformOp	createTransform(String name, INDArray x, INDArray y)
TransformOp	createTransform(String name, INDArray x, INDArray y, INDArray z)

Method Detail

createLossFunction

LossFunction createLossFunction(String name,
                                INDArray x,
                                INDArray y)

Create a loss function with the given inputs and outputs

Parameters:

name - the name of the function

x - the input

y - the output

Returns:

a loss function representing the delta between the 2

createAccum

Accumulation createAccum(String name,
                         INDArray x)

Accumulation operation

Parameters:

name - the name of the function to create

x - the input to the function

Returns:

the operation

createAccum

Accumulation createAccum(String name,
                         INDArray x,
                         INDArray y,
                         INDArray z)

Accumulation operation

Parameters:

name - the name of the function

x - the input

y - the pairwise transformation

z - the output

Returns:

the operation

createAccum

Accumulation createAccum(String name,
                         INDArray x,
                         INDArray y)

Parameters:

name -

x -

y -

Returns:

createIndexAccum

IndexAccumulation createIndexAccum(String name,
                                   INDArray x)

Index accumulation operation

Parameters:

name - the name of the function to create

x - the input to the function

Returns:

the operation

createIndexAccum

IndexAccumulation createIndexAccum(String name,
                                   INDArray x,
                                   INDArray y)

Index accumulation operation

Parameters:

name -

x -

y -

Returns:

createTransform

TransformOp createTransform(String name,
                            INDArray x,
                            INDArray y)

Parameters:

name -

x -

y -

Returns:

createTransform

TransformOp createTransform(String name,
                            INDArray x)

Parameters:

name -

x -

Returns:

createTransform

TransformOp createTransform(String name,
                            INDArray x,
                            INDArray y,
                            INDArray z)

Parameters:

name -

x -

y -

z -

Returns:

createBroadcastOp

BroadcastOp createBroadcastOp(String name,
                              INDArray x,
                              INDArray y,
                              INDArray z,
                              int... dimension)

Create a vector operation

Parameters:

name - Name of the vector op

x - NDArray to operate on

y - Vector

z - Result NDArray

dimension - Dimension to do op along. 0 for row, 1 for column, etc

Returns:

VectorOp

createBroadcastOp

BroadcastOp createBroadcastOp(String name,
                              INDArray x,
                              INDArray z,
                              int... dimension)

Create a vector operation

Parameters:

name - Name of the vector op

x - NDArray to operate on

z - Result NDArray

dimension - Dimension to do op along. 0 for row, 1 for column, etc

Returns:

VectorOp