org.nd4j.autodiff.samediff.ops

Class SDLinalg

java.lang.Object

org.nd4j.autodiff.samediff.ops.SDOps

org.nd4j.autodiff.samediff.ops.SDLinalg

public class SDLinalg
extends SDOps

Field Summary

Fields inherited from class org.nd4j.autodiff.samediff.ops.SDOps

sd

Constructor Summary

Constructors

Constructor and Description
SDLinalg(SameDiff sameDiff)

Method Summary

Modifier and Type	Method and Description
SDVariable	cholesky(SDVariable input) Computes the Cholesky decomposition of one or more square matrices.
SDVariable	cholesky(String name, SDVariable input) Computes the Cholesky decomposition of one or more square matrices.
SDVariable	cross(SDVariable a, SDVariable b) Computes pairwise cross product.
SDVariable	cross(String name, SDVariable a, SDVariable b) Computes pairwise cross product.
SDVariable	diag_part(SDVariable input) Calculates diagonal tensor.
SDVariable	diag_part(String name, SDVariable input) Calculates diagonal tensor.
SDVariable	diag(SDVariable input) Calculates diagonal tensor.
SDVariable	diag(String name, SDVariable input) Calculates diagonal tensor.
SDVariable	logdet(SDVariable input) Calculates log of determinant.
SDVariable	logdet(String name, SDVariable input) Calculates log of determinant.
SDVariable	lstsq(SDVariable matrix, SDVariable rhs, double l2_reguralizer) Solver for linear squares problems.
SDVariable	lstsq(SDVariable matrix, SDVariable rhs, double l2_reguralizer, boolean fast) Solver for linear squares problems.
SDVariable	lstsq(String name, SDVariable matrix, SDVariable rhs, double l2_reguralizer) Solver for linear squares problems.
SDVariable	lstsq(String name, SDVariable matrix, SDVariable rhs, double l2_reguralizer, boolean fast) Solver for linear squares problems.
SDVariable	lu(SDVariable input) Computes LU decomposition.
SDVariable	lu(String name, SDVariable input) Computes LU decomposition.
SDVariable	matmul(SDVariable a, SDVariable b) Performs matrix mutiplication on input tensors.
SDVariable	matmul(String name, SDVariable a, SDVariable b) Performs matrix mutiplication on input tensors.
SDVariable[]	matrixBandPart(SDVariable input, int minLower, int maxUpper) Copy a tensor setting outside a central band in each innermost matrix.
SDVariable[]	matrixBandPart(String[] names, SDVariable input, int minLower, int maxUpper) Copy a tensor setting outside a central band in each innermost matrix.
SDVariable	mmul(SDVariable x, SDVariable y) Matrix multiplication: out = mmul(x,y) Supports specifying transpose argument to perform operation such as mmul(a^T, b), etc.
SDVariable	mmul(SDVariable x, SDVariable y, boolean transposeX, boolean transposeY, boolean transposeZ) Matrix multiplication: out = mmul(x,y) Supports specifying transpose argument to perform operation such as mmul(a^T, b), etc.
SDVariable	mmul(String name, SDVariable x, SDVariable y) Matrix multiplication: out = mmul(x,y) Supports specifying transpose argument to perform operation such as mmul(a^T, b), etc.
SDVariable	mmul(String name, SDVariable x, SDVariable y, boolean transposeX, boolean transposeY, boolean transposeZ) Matrix multiplication: out = mmul(x,y) Supports specifying transpose argument to perform operation such as mmul(a^T, b), etc.
SDVariable[]	qr(SDVariable input) Computes the QR decompositions of input matrix.
SDVariable[]	qr(SDVariable input, boolean full) Computes the QR decompositions of input matrix.
SDVariable[]	qr(String[] names, SDVariable input) Computes the QR decompositions of input matrix.
SDVariable[]	qr(String[] names, SDVariable input, boolean full) Computes the QR decompositions of input matrix.
SDVariable	solve(SDVariable matrix, SDVariable rhs) Solver for systems of linear equations.
SDVariable	solve(SDVariable matrix, SDVariable rhs, boolean adjoint) Solver for systems of linear equations.
SDVariable	solve(String name, SDVariable matrix, SDVariable rhs) Solver for systems of linear equations.
SDVariable	solve(String name, SDVariable matrix, SDVariable rhs, boolean adjoint) Solver for systems of linear equations.
SDVariable	svd(SDVariable input, boolean fullUV, boolean computeUV) Calculates singular value decomposition.
SDVariable	svd(SDVariable input, boolean fullUV, boolean computeUV, int switchNum) Calculates singular value decomposition.
SDVariable	svd(String name, SDVariable input, boolean fullUV, boolean computeUV) Calculates singular value decomposition.
SDVariable	svd(String name, SDVariable input, boolean fullUV, boolean computeUV, int switchNum) Calculates singular value decomposition.
SDVariable	tri(DataType dataType, int row, int column, int diagonal) An array with ones at and below the given diagonal and zeros elsewhere.
SDVariable	tri(int row, int column) An array with ones at and below the given diagonal and zeros elsewhere.
SDVariable	tri(String name, DataType dataType, int row, int column, int diagonal) An array with ones at and below the given diagonal and zeros elsewhere.
SDVariable	tri(String name, int row, int column) An array with ones at and below the given diagonal and zeros elsewhere.
SDVariable	triangularSolve(SDVariable matrix, SDVariable rhs, boolean lower, boolean adjoint) Solver for systems of linear questions.
SDVariable	triangularSolve(String name, SDVariable matrix, SDVariable rhs, boolean lower, boolean adjoint) Solver for systems of linear questions.
SDVariable	triu(SDVariable input) Upper triangle of an array.
SDVariable	triu(SDVariable input, int diag) Upper triangle of an array.
SDVariable	triu(String name, SDVariable input) Upper triangle of an array.
SDVariable	triu(String name, SDVariable input, int diag) Upper triangle of an array.

Methods inherited from class java.lang.Object

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

Constructor Detail

SDLinalg

public SDLinalg(SameDiff sameDiff)

Method Detail

cholesky

public SDVariable cholesky(SDVariable input)

Computes the Cholesky decomposition of one or more square matrices.

Parameters:

input - Input tensor with inner-most 2 dimensions forming square matrices (NUMERIC type)

Returns:

output Transformed tensor (NUMERIC type)

cholesky

public SDVariable cholesky(String name,
                           SDVariable input)

Computes the Cholesky decomposition of one or more square matrices.

Parameters:

name - name May be null. Name for the output variable

input - Input tensor with inner-most 2 dimensions forming square matrices (NUMERIC type)

Returns:

output Transformed tensor (NUMERIC type)

lstsq

public SDVariable lstsq(SDVariable matrix,
                        SDVariable rhs,
                        double l2_reguralizer,
                        boolean fast)

Solver for linear squares problems.

Parameters:

matrix - input tensor (NUMERIC type)

rhs - input tensor (NUMERIC type)

l2_reguralizer - regularizer

fast - fast mode, defaults to True

Returns:

output Transformed tensor (FLOATING_POINT type)

lstsq

public SDVariable lstsq(String name,
                        SDVariable matrix,
                        SDVariable rhs,
                        double l2_reguralizer,
                        boolean fast)

Solver for linear squares problems.

Parameters:

name - name May be null. Name for the output variable

matrix - input tensor (NUMERIC type)

rhs - input tensor (NUMERIC type)

l2_reguralizer - regularizer

fast - fast mode, defaults to True

Returns:

output Transformed tensor (FLOATING_POINT type)

lstsq

public SDVariable lstsq(SDVariable matrix,
                        SDVariable rhs,
                        double l2_reguralizer)

Solver for linear squares problems.

Parameters:

matrix - input tensor (NUMERIC type)

rhs - input tensor (NUMERIC type)

l2_reguralizer - regularizer

Returns:

output Transformed tensor (FLOATING_POINT type)

lstsq

public SDVariable lstsq(String name,
                        SDVariable matrix,
                        SDVariable rhs,
                        double l2_reguralizer)

Solver for linear squares problems.

Parameters:

name - name May be null. Name for the output variable

matrix - input tensor (NUMERIC type)

rhs - input tensor (NUMERIC type)

l2_reguralizer - regularizer

Returns:

output Transformed tensor (FLOATING_POINT type)

lu

public SDVariable lu(SDVariable input)

Computes LU decomposition.

Parameters:

input - input tensor (NUMERIC type)

Returns:

output (FLOATING_POINT type)

lu

public SDVariable lu(String name,
                     SDVariable input)

Computes LU decomposition.

Parameters:

name - name May be null. Name for the output variable

input - input tensor (NUMERIC type)

Returns:

output (FLOATING_POINT type)

matmul

public SDVariable matmul(SDVariable a,
                         SDVariable b)

Performs matrix mutiplication on input tensors.

Parameters:

a - input tensor (NUMERIC type)

b - input tensor (NUMERIC type)

Returns:

output (FLOATING_POINT type)

matmul

public SDVariable matmul(String name,
                         SDVariable a,
                         SDVariable b)

Performs matrix mutiplication on input tensors.

Parameters:

name - name May be null. Name for the output variable

a - input tensor (NUMERIC type)

b - input tensor (NUMERIC type)

Returns:

output (FLOATING_POINT type)

matrixBandPart

public SDVariable[] matrixBandPart(SDVariable input,
                                   int minLower,
                                   int maxUpper)

Copy a tensor setting outside a central band in each innermost matrix.

Parameters:

input - input tensor (NUMERIC type)

minLower - lower diagonal count

maxUpper - upper diagonal count

matrixBandPart

public SDVariable[] matrixBandPart(String[] names,
                                   SDVariable input,
                                   int minLower,
                                   int maxUpper)

Copy a tensor setting outside a central band in each innermost matrix.

Parameters:

names - names May be null. Arrays of names for the output variables.

input - input tensor (NUMERIC type)

minLower - lower diagonal count

maxUpper - upper diagonal count

qr

public SDVariable[] qr(SDVariable input,
                       boolean full)

Computes the QR decompositions of input matrix.

Parameters:

input - input tensor (NUMERIC type)

full - full matrices mode

qr

public SDVariable[] qr(String[] names,
                       SDVariable input,
                       boolean full)

Computes the QR decompositions of input matrix.

Parameters:

names - names May be null. Arrays of names for the output variables.

input - input tensor (NUMERIC type)

full - full matrices mode

qr

public SDVariable[] qr(SDVariable input)

Computes the QR decompositions of input matrix.

Parameters:

input - input tensor (NUMERIC type)

qr

public SDVariable[] qr(String[] names,
                       SDVariable input)

Computes the QR decompositions of input matrix.

Parameters:

names - names May be null. Arrays of names for the output variables.

input - input tensor (NUMERIC type)

solve

public SDVariable solve(SDVariable matrix,
                        SDVariable rhs,
                        boolean adjoint)

Solver for systems of linear equations.

Parameters:

matrix - input tensor (NUMERIC type)

rhs - input tensor (NUMERIC type)

adjoint - adjoint mode, defaults to False

Returns:

output Output tensor (FLOATING_POINT type)

solve

public SDVariable solve(String name,
                        SDVariable matrix,
                        SDVariable rhs,
                        boolean adjoint)

Solver for systems of linear equations.

Parameters:

name - name May be null. Name for the output variable

matrix - input tensor (NUMERIC type)

rhs - input tensor (NUMERIC type)

adjoint - adjoint mode, defaults to False

Returns:

output Output tensor (FLOATING_POINT type)

solve

public SDVariable solve(SDVariable matrix,
                        SDVariable rhs)

Solver for systems of linear equations.

Parameters:

matrix - input tensor (NUMERIC type)

rhs - input tensor (NUMERIC type)

Returns:

output Output tensor (FLOATING_POINT type)

solve

public SDVariable solve(String name,
                        SDVariable matrix,
                        SDVariable rhs)

Solver for systems of linear equations.

Parameters:

name - name May be null. Name for the output variable

matrix - input tensor (NUMERIC type)

rhs - input tensor (NUMERIC type)

Returns:

output Output tensor (FLOATING_POINT type)

triangularSolve

public SDVariable triangularSolve(SDVariable matrix,
                                  SDVariable rhs,
                                  boolean lower,
                                  boolean adjoint)

Solver for systems of linear questions.

Parameters:

matrix - input tensor (NUMERIC type)

rhs - input tensor (NUMERIC type)

lower - defines whether innermost matrices in matrix are lower or upper triangular

adjoint - adjoint mode

Returns:

output (FLOATING_POINT type)

triangularSolve

public SDVariable triangularSolve(String name,
                                  SDVariable matrix,
                                  SDVariable rhs,
                                  boolean lower,
                                  boolean adjoint)

Solver for systems of linear questions.

Parameters:

name - name May be null. Name for the output variable

matrix - input tensor (NUMERIC type)

rhs - input tensor (NUMERIC type)

lower - defines whether innermost matrices in matrix are lower or upper triangular

adjoint - adjoint mode

Returns:

output (FLOATING_POINT type)

cross

public SDVariable cross(SDVariable a,
                        SDVariable b)

Computes pairwise cross product.

Parameters:

a - (NUMERIC type)

b - (NUMERIC type)

Returns:

output (FLOATING_POINT type)

cross

public SDVariable cross(String name,
                        SDVariable a,
                        SDVariable b)

Computes pairwise cross product.

Parameters:

name - name May be null. Name for the output variable

a - (NUMERIC type)

b - (NUMERIC type)

Returns:

output (FLOATING_POINT type)

diag

public SDVariable diag(SDVariable input)

Calculates diagonal tensor.

Parameters:

input - (NUMERIC type)

Returns:

output (FLOATING_POINT type)

diag

public SDVariable diag(String name,
                       SDVariable input)

Calculates diagonal tensor.

Parameters:

name - name May be null. Name for the output variable

input - (NUMERIC type)

Returns:

output (FLOATING_POINT type)

diag_part

public SDVariable diag_part(SDVariable input)

Calculates diagonal tensor.

Parameters:

input - (NUMERIC type)

Returns:

output (FLOATING_POINT type)

diag_part

public SDVariable diag_part(String name,
                            SDVariable input)

Calculates diagonal tensor.

Parameters:

name - name May be null. Name for the output variable

input - (NUMERIC type)

Returns:

output (FLOATING_POINT type)

logdet

public SDVariable logdet(SDVariable input)

Calculates log of determinant.

Parameters:

input - (NUMERIC type)

Returns:

output (FLOATING_POINT type)

logdet

public SDVariable logdet(String name,
                         SDVariable input)

Calculates log of determinant.

Parameters:

name - name May be null. Name for the output variable

input - (NUMERIC type)

Returns:

output (FLOATING_POINT type)

mmul

public SDVariable mmul(SDVariable x,
                       SDVariable y,
                       boolean transposeX,
                       boolean transposeY,
                       boolean transposeZ)

Matrix multiplication: out = mmul(x,y)
Supports specifying transpose argument to perform operation such as mmul(a^T, b), etc.

Parameters:

x - First input variable (NUMERIC type)

y - Second input variable (NUMERIC type)

transposeX - Transpose x (first argument)

transposeY - Transpose y (second argument)

transposeZ - Transpose result array

Returns:

output (NUMERIC type)

mmul

public SDVariable mmul(String name,
                       SDVariable x,
                       SDVariable y,
                       boolean transposeX,
                       boolean transposeY,
                       boolean transposeZ)

Matrix multiplication: out = mmul(x,y)
Supports specifying transpose argument to perform operation such as mmul(a^T, b), etc.

Parameters:

name - name May be null. Name for the output variable

x - First input variable (NUMERIC type)

y - Second input variable (NUMERIC type)

transposeX - Transpose x (first argument)

transposeY - Transpose y (second argument)

transposeZ - Transpose result array

Returns:

output (NUMERIC type)

mmul

public SDVariable mmul(SDVariable x,
                       SDVariable y)

Matrix multiplication: out = mmul(x,y)
Supports specifying transpose argument to perform operation such as mmul(a^T, b), etc.

Parameters:

x - First input variable (NUMERIC type)

y - Second input variable (NUMERIC type)

Returns:

output (NUMERIC type)

mmul

public SDVariable mmul(String name,
                       SDVariable x,
                       SDVariable y)

Matrix multiplication: out = mmul(x,y)
Supports specifying transpose argument to perform operation such as mmul(a^T, b), etc.

Parameters:

name - name May be null. Name for the output variable

x - First input variable (NUMERIC type)

y - Second input variable (NUMERIC type)

Returns:

output (NUMERIC type)

svd

public SDVariable svd(SDVariable input,
                      boolean fullUV,
                      boolean computeUV,
                      int switchNum)

Calculates singular value decomposition.

Parameters:

input - (NUMERIC type)

fullUV -

computeUV -

switchNum -

Returns:

output (FLOATING_POINT type)

svd

public SDVariable svd(String name,
                      SDVariable input,
                      boolean fullUV,
                      boolean computeUV,
                      int switchNum)

Calculates singular value decomposition.

Parameters:

name - name May be null. Name for the output variable

input - (NUMERIC type)

fullUV -

computeUV -

switchNum -

Returns:

output (FLOATING_POINT type)

svd

public SDVariable svd(SDVariable input,
                      boolean fullUV,
                      boolean computeUV)

Calculates singular value decomposition.

Parameters:

input - (NUMERIC type)

fullUV -

computeUV -

Returns:

output (FLOATING_POINT type)

svd

public SDVariable svd(String name,
                      SDVariable input,
                      boolean fullUV,
                      boolean computeUV)

Calculates singular value decomposition.

Parameters:

name - name May be null. Name for the output variable

input - (NUMERIC type)

fullUV -

computeUV -

Returns:

output (FLOATING_POINT type)

tri

public SDVariable tri(DataType dataType,
                      int row,
                      int column,
                      int diagonal)

An array with ones at and below the given diagonal and zeros elsewhere.

Parameters:

dataType - Data type

row -

column -

diagonal -

Returns:

output (FLOATING_POINT type)

tri

public SDVariable tri(String name,
                      DataType dataType,
                      int row,
                      int column,
                      int diagonal)

An array with ones at and below the given diagonal and zeros elsewhere.

Parameters:

name - name May be null. Name for the output variable

dataType - Data type

row -

column -

diagonal -

Returns:

output (FLOATING_POINT type)

tri

public SDVariable tri(int row,
                      int column)

An array with ones at and below the given diagonal and zeros elsewhere.

Parameters:

row -

column -

Returns:

output (FLOATING_POINT type)

tri

public SDVariable tri(String name,
                      int row,
                      int column)

An array with ones at and below the given diagonal and zeros elsewhere.

Parameters:

name - name May be null. Name for the output variable

row -

column -

Returns:

output (FLOATING_POINT type)

triu

public SDVariable triu(SDVariable input,
                       int diag)

Upper triangle of an array. Return a copy of a input tensor with the elements below the k-th diagonal zeroed.

Parameters:

input - (NUMERIC type)

diag -

Returns:

output (FLOATING_POINT type)

triu

public SDVariable triu(String name,
                       SDVariable input,
                       int diag)

Upper triangle of an array. Return a copy of a input tensor with the elements below the k-th diagonal zeroed.

Parameters:

name - name May be null. Name for the output variable

input - (NUMERIC type)

diag -

Returns:

output (FLOATING_POINT type)

triu

public SDVariable triu(SDVariable input)

Upper triangle of an array. Return a copy of a input tensor with the elements below the k-th diagonal zeroed.

Parameters:

input - (NUMERIC type)

Returns:

output (FLOATING_POINT type)

triu

public SDVariable triu(String name,
                       SDVariable input)

Upper triangle of an array. Return a copy of a input tensor with the elements below the k-th diagonal zeroed.

Parameters:

name - name May be null. Name for the output variable

input - (NUMERIC type)

Returns:

output (FLOATING_POINT type)