tensorflow :: operaciones :: MatrixDiagV3
#include <array_ops.h>Devuelve un tensor diagonal por lotes con valores diagonales por lotes dados.
Resumen
Devuelve un tensor con el contenido en diagonal como diagonal k[0] a k[1] -ésima de una matriz, con todo lo demás relleno con padding . num_rows y num_cols especifican la dimensión de la matriz más interna de la salida. Si no se especifican ambos, la operación asume que la matriz más interna es cuadrada e infiere su tamaño de k y la dimensión más interna de la diagonal . Si solo se especifica uno de ellos, la operación asume que el valor no especificado es el más pequeño posible según otros criterios.
Deje que la diagonal tenga r dimensiones [I, J, ..., L, M, N] . El tensor de salida tiene rango r+1 con forma [I, J, ..., L, M, num_rows, num_cols] cuando solo se da una diagonal ( k es un número entero o k[0] == k[1] ) . De lo contrario, tiene rango r con forma [I, J, ..., L, num_rows, num_cols] .
La segunda dimensión más interna de la diagonal tiene un doble significado. Cuando k es escalar o k[0] == k[1] , M es parte del tamaño del lote [I, J, ..., M] y el tensor de salida es:
output[i, j, ..., l, m, n]
= diagonal[i, j, ..., l, n-max(d_upper, 0)] ; if n - m == d_upper
padding_value ; otherwise
De lo contrario, M se trata como el número de diagonales de la matriz en el mismo lote ( M = k[1]-k[0]+1 ) y el tensor de salida es:
output[i, j, ..., l, m, n]
= diagonal[i, j, ..., l, diag_index, index_in_diag] ; if k[0] <= d <= k[1]
padding_value ; otherwise
d = n - m , diag_index = [k] - d e index_in_diag = n - max(d, 0) + offset . offset es cero excepto cuando la alineación de la diagonal es hacia la derecha.
offset = max_diag_len - diag_len(d) ; if (`align` in {RIGHT_LEFT, RIGHT_RIGHT}
and `d >= 0`) or
(`align` in {LEFT_RIGHT, RIGHT_RIGHT}
and `d <= 0`)
0 ; otherwise
diag_len(d) = min(cols - max(d, 0), rows + min(d, 0)) .Por ejemplo:
# The main diagonal.
diagonal = np.array([[1, 2, 3, 4], # Input shape: (2, 4)
[5, 6, 7, 8]])
tf.matrix_diag(diagonal) ==> [[[1, 0, 0, 0], # Output shape: (2, 4, 4)
[0, 2, 0, 0],
[0, 0, 3, 0],
[0, 0, 0, 4]],
[[5, 0, 0, 0],
[0, 6, 0, 0],
[0, 0, 7, 0],
[0, 0, 0, 8]]]
# A superdiagonal (per batch).
diagonal = np.array([[1, 2, 3], # Input shape: (2, 3)
[4, 5, 6]])
tf.matrix_diag(diagonal, k = 1)
==> [[[0, 1, 0, 0], # Output shape: (2, 4, 4)
[0, 0, 2, 0],
[0, 0, 0, 3],
[0, 0, 0, 0]],
[[0, 4, 0, 0],
[0, 0, 5, 0],
[0, 0, 0, 6],
[0, 0, 0, 0]]]
# A tridiagonal band (per batch).
diagonals = np.array([[[0, 8, 9], # Input shape: (2, 2, 3)
[1, 2, 3],
[4, 5, 0]],
[[0, 2, 3],
[6, 7, 9],
[9, 1, 0]]])
tf.matrix_diag(diagonals, k = (-1, 1))
==> [[[1, 8, 0], # Output shape: (2, 3, 3)
[4, 2, 9],
[0, 5, 3]],
[[6, 2, 0],
[9, 7, 3],
[0, 1, 9]]]
# LEFT_RIGHT alignment.
diagonals = np.array([[[8, 9, 0], # Input shape: (2, 2, 3)
[1, 2, 3],
[0, 4, 5]],
[[2, 3, 0],
[6, 7, 9],
[0, 9, 1]]])
tf.matrix_diag(diagonals, k = (-1, 1), align="LEFT_RIGHT")
==> [[[1, 8, 0], # Output shape: (2, 3, 3)
[4, 2, 9],
[0, 5, 3]],
[[6, 2, 0],
[9, 7, 3],
[0, 1, 9]]]
# Rectangular matrix.
diagonal = np.array([1, 2]) # Input shape: (2)
tf.matrix_diag(diagonal, k = -1, num_rows = 3, num_cols = 4)
==> [[0, 0, 0, 0], # Output shape: (3, 4)
[1, 0, 0, 0],
[0, 2, 0, 0]]
# Rectangular matrix with inferred num_cols and padding_value = 9.
tf.matrix_diag(diagonal, k = -1, num_rows = 3, padding_value = 9)
==> [[9, 9], # Output shape: (3, 2)
[1, 9],
[9, 2]]
Arguments:
- scope: A Scope object
- diagonal: Rank
r, wherer >= 1 - k: Diagonal offset(s). Positive value means superdiagonal, 0 refers to the main diagonal, and negative value means subdiagonals.
kcan be a single integer (for a single diagonal) or a pair of integers specifying the low and high ends of a matrix band.k[0]must not be larger thank[1]. - num_rows: The number of rows of the output matrix. If it is not provided, the op assumes the output matrix is a square matrix and infers the matrix size from k and the innermost dimension of
diagonal. - num_cols: The number of columns of the output matrix. If it is not provided, the op assumes the output matrix is a square matrix and infers the matrix size from k and the innermost dimension of
diagonal. - padding_value: The number to fill the area outside the specified diagonal band with. Default is 0.
Optional attributes (see Attrs):
- align: Some diagonals are shorter than
max_diag_lenand need to be padded.alignis a string specifying how superdiagonals and subdiagonals should be aligned, respectively. There are four possible alignments: "RIGHT_LEFT" (default), "LEFT_RIGHT", "LEFT_LEFT", and "RIGHT_RIGHT". "RIGHT_LEFT" aligns superdiagonals to the right (left-pads the row) and subdiagonals to the left (right-pads the row). It is the packing format LAPACK uses. cuSPARSE uses "LEFT_RIGHT", which is the opposite alignment.
Returns:
Output: Has rankr+1whenkis an integer ork[0] == k[1], rankrotherwise.
Constructors and Destructors | |
|---|---|
MatrixDiagV3(const ::tensorflow::Scope & scope, ::tensorflow::Input diagonal, ::tensorflow::Input k, ::tensorflow::Input num_rows, ::tensorflow::Input num_cols, ::tensorflow::Input padding_value) | |
MatrixDiagV3(const ::tensorflow::Scope & scope, ::tensorflow::Input diagonal, ::tensorflow::Input k, ::tensorflow::Input num_rows, ::tensorflow::Input num_cols, ::tensorflow::Input padding_value, const MatrixDiagV3::Attrs & attrs) |
Public functions | |
|---|---|
node() const | ::tensorflow::Node * |
operator::tensorflow::Input() const | |
operator::tensorflow::Output() const | |
Structs | |
|---|---|
tensorflow:: | Optional attribute setters for MatrixDiagV3. |