ScatterNd

Scatter `updates` into a new tensor according to `indices`.

Creates a new tensor by applying sparse `updates` to individual values or slices within a tensor (initially zero for numeric, empty for string) of the given `shape` according to indices. This operator is the inverse of the tf.gather_nd operator which extracts values or slices from a given tensor.

This operation is similar to tensor_scatter_add, except that the tensor is zero-initialized. Calling tf.scatter_nd(indices, values, shape) is identical to `tensor_scatter_add(tf.zeros(shape, values.dtype), indices, values)`

If `indices` contains duplicates, then their updates are accumulated (summed).

WARNING: The order in which updates are applied is nondeterministic, so the output will be nondeterministic if `indices` contains duplicates -- because of some numerical approximation issues, numbers summed in different order may yield different results.

`indices` is an integer tensor containing indices into a new tensor of shape `shape`. The last dimension of `indices` can be at most the rank of `shape`:

indices.shape[-1] <= shape.rank

The last dimension of `indices` corresponds to indices into elements (if `indices.shape[-1] = shape.rank`) or slices (if `indices.shape[-1] < shape.rank`) along dimension `indices.shape[-1]` of `shape`. `updates` is a tensor with shape

indices.shape[:-1] + shape[indices.shape[-1]:]

The simplest form of scatter is to insert individual elements in a tensor by index. For example, say we want to insert 4 scattered elements in a rank-1 tensor with 8 elements.

In Python, this scatter operation would look like this:

indices = tf.constant([[4], [3], [1], [7]])
     updates = tf.constant([9, 10, 11, 12])
     shape = tf.constant([8])
     scatter = tf.scatter_nd(indices, updates, shape)
     print(scatter)
 

The resulting tensor would look like this:

[0, 11, 0, 10, 9, 0, 0, 12]

We can also, insert entire slices of a higher rank tensor all at once. For example, if we wanted to insert two slices in the first dimension of a rank-3 tensor with two matrices of new values.

In Python, this scatter operation would look like this:

indices = tf.constant([[0], [2]])
     updates = tf.constant([[[5, 5, 5, 5], [6, 6, 6, 6],
                             [7, 7, 7, 7], [8, 8, 8, 8]],
                            [[5, 5, 5, 5], [6, 6, 6, 6],
                             [7, 7, 7, 7], [8, 8, 8, 8]]])
     shape = tf.constant([4, 4, 4])
     scatter = tf.scatter_nd(indices, updates, shape)
     print(scatter)
 

The resulting tensor would look like this:

[[[5, 5, 5, 5], [6, 6, 6, 6], [7, 7, 7, 7], [8, 8, 8, 8]], [[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]], [[5, 5, 5, 5], [6, 6, 6, 6], [7, 7, 7, 7], [8, 8, 8, 8]], [[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]]]

Note that on CPU, if an out of bound index is found, an error is returned. On GPU, if an out of bound index is found, the index is ignored.

Constants

String

OP_NAME

The name of this op, as known by TensorFlow core engine

Public Methods

Output<U>	asOutput() Returns the symbolic handle of the tensor.
static <U extends TType, T extends TNumber> ScatterNd<U>	create(Scope scope, Operand<T> indices, Operand<U> updates, Operand<T> shape) Factory method to create a class wrapping a new ScatterNd operation.
Output<U>	output() A new tensor with the given shape and updates applied according to the indices.

Inherited Methods

From class org.tensorflow.op.RawOp

final boolean	equals(Object obj)
final int	hashCode()
Operation	op() Return this unit of computation as a single Operation.
final String	toString()

From class java.lang.Object

boolean	equals(Object arg0)
final Class<?>	getClass()
int	hashCode()
final void	notify()
final void	notifyAll()
String	toString()
final void	wait(long arg0, int arg1)
final void	wait(long arg0)
final void	wait()

From interface org.tensorflow.op.Op

abstract ExecutionEnvironment	env() Return the execution environment this op was created in.
abstract Operation	op() Return this unit of computation as a single Operation.

From interface org.tensorflow.Operand

abstract Output<U extends TType>	asOutput() Returns the symbolic handle of the tensor.
abstract U extends TType	asTensor() Returns the tensor at this operand.
abstract Shape	shape() Returns the (possibly partially known) shape of the tensor referred to by the Output of this operand.
abstract Class<U extends TType>	type() Returns the tensor type of this operand

From interface org.tensorflow.ndarray.Shaped

abstract int	rank()
abstract Shape	shape()
abstract long	size() Computes and returns the total size of this container, in number of values.