T - data type for band() output@Operator public final class MatrixBandPart<T> extends PrimitiveOp implements Operand<T>
to zero.
The `band` part is computed as follows: Assume `input` has `k` dimensions `[I, J, K, ..., M, N]`, then the output is a tensor with the same shape where
`band[i, j, k, ..., m, n] = in_band(m, n) * input[i, j, k, ..., m, n]`.
The indicator function
`in_band(m, n) = (num_lower < 0 || (m-n) <= num_lower)) && (num_upper < 0 || (n-m) <= num_upper)`.
For example:
# if 'input' is [[ 0, 1, 2, 3]
[-1, 0, 1, 2]
[-2, -1, 0, 1]
[-3, -2, -1, 0]],
tf.matrix_band_part(input, 1, -1) ==> [[ 0, 1, 2, 3]
[-1, 0, 1, 2]
[ 0, -1, 0, 1]
[ 0, 0, -1, 0]],
tf.matrix_band_part(input, 2, 1) ==> [[ 0, 1, 0, 0]
[-1, 0, 1, 0]
[-2, -1, 0, 1]
[ 0, -2, -1, 0]]
Useful special cases:
tf.matrix_band_part(input, 0, -1) ==> Upper triangular part.
tf.matrix_band_part(input, -1, 0) ==> Lower triangular part.
tf.matrix_band_part(input, 0, 0) ==> Diagonal.
operation| Modifier and Type | Method and Description |
|---|---|
Output<T> |
asOutput()
Returns the symbolic handle of a tensor.
|
Output<T> |
band()
Rank `k` tensor of the same shape as input.
|
static <T,U extends Number> |
create(Scope scope,
Operand<T> input,
Operand<U> numLower,
Operand<U> numUpper)
Factory method to create a class to wrap a new MatrixBandPart operation to the graph.
|
equals, hashCode, toStringpublic static <T,U extends Number> MatrixBandPart<T> create(Scope scope, Operand<T> input, Operand<U> numLower, Operand<U> numUpper)
scope - current graph scopeinput - Rank `k` tensor.numLower - 0-D tensor. Number of subdiagonals to keep. If negative, keep entire
lower triangle.numUpper - 0-D tensor. Number of superdiagonals to keep. If negative, keep
entire upper triangle.public Output<T> band()
public Output<T> asOutput()
OperandInputs to TensorFlow operations are outputs of another TensorFlow operation. This method is used to obtain a symbolic handle that represents the computation of the input.
asOutput in interface Operand<T>OperationBuilder.addInput(Output)Copyright © 2015–2019. All rights reserved.