PermutationMatrix¶

class tfsnippet.utils.PermutationMatrix(data)¶

A non-trainable permutation matrix.

Attributes Summary

`col_permutation`	Get the column permutation indices.
`row_permutation`	Get the row permutation indices.
`shape`	Get the shape of this permutation matrix.

Methods Summary

`det`()	Get the determinant of this permutation matrix.
`get_numpy_matrix`([dtype])	Get the numpy permutation matrix.
`inv`()	Get the inverse permutation matrix of this matrix.
`left_mult`(input[, name])	Left multiply to input matrix.
`right_mult`(input[, name])	Right multiply to input matrix.

Attributes Documentation

col_permutation¶

Get the column permutation indices.

Returns:	The column permutation indices.
Return type:	tuple[int]

row_permutation¶

Get the row permutation indices.

Returns:	The row permutation indices.
Return type:	tuple[int]

shape¶

Get the shape of this permutation matrix.

Returns:	The shape of this permutation matrix.
Return type:	(int, int)

Methods Documentation

det()¶

Get the determinant of this permutation matrix.

Returns:	The determinant of this permutation matrix.
Return type:	float

get_numpy_matrix(dtype=<type 'numpy.int32'>)¶

Get the numpy permutation matrix.

Parameters:	dtype – The data type of the returned matrix.
Returns:	A 2-d numpy matrix.
Return type:	np.ndarray

inv()¶

Get the inverse permutation matrix of this matrix.

Returns:	The inverse permutation matrix.
Return type:	PermutationMatrix

left_mult(input, name=None)¶

Left multiply to input matrix.

output = matmul(self, input)

Parameters:	input (np.ndarray or tf.Tensor) – The input matrix, whose shape must be `(self.shape[1], ?)`. name (str) – Default name of the name scope. If not specified, generate one according to the method name.
Returns:	The result of multiplication.
Return type:	np.ndarray or tf.Tensor

right_mult(input, name=None)¶

Right multiply to input matrix.

output = matmul(input, self)

Parameters:	input (np.ndarray or tf.Tensor) – The input matrix, whose shape must be `(?, self.shape[0])`. name (str) – Default name of the name scope. If not specified, generate one according to the method name.
Returns:	The result of multiplication.
Return type:	np.ndarray or tf.Tensor