View source on GitHub |
Count the number of times an integer value appears in a tensor.
tf.sparse.bincount(
values,
weights=None,
axis=0,
minlength=None,
maxlength=None,
binary_output=False,
name=None
)
This op takes an N-dimensional Tensor
, RaggedTensor
, or SparseTensor
,
and returns an N-dimensional int64 SparseTensor where element
[i0...i[axis], j]
contains the number of times the value j
appears in
slice [i0...i[axis], :]
of the input tensor. Currently, only N=0 and
N=-1 are supported.
Returns | |
---|---|
A SparseTensor with output.shape = values.shape[:axis] + [N] , where N is
|
Raises | |
---|---|
InvalidArgumentError if negative values are provided as an input.
|
Examples:
Bin-counting every item in individual batches
This example takes an input (which could be a Tensor, RaggedTensor, or SparseTensor) and returns a SparseTensor where the value of (i,j) is the number of times value j appears in batch i.
data = np.array([[10, 20, 30, 20], [11, 101, 11, 10001]], dtype=np.int64)
tf.sparse.bincount(data, axis=-1)
SparseTensor(indices=tf.Tensor(
[[ 0 10]
[ 0 20]
[ 0 30]
[ 1 11]
[ 1 101]
[ 1 10001]], shape=(6, 2), dtype=int64),
values=tf.Tensor([1 2 1 2 1 1], shape=(6,), dtype=int64),
dense_shape=tf.Tensor([ 2 10002], shape=(2,), dtype=int64))
This example shows a sparse tensor input. Missing zeros are not counted.
data = tf.sparse.SparseTensor(
indices=[[0, 3], [0, 7], [0, 8], [0, 11],
[1, 9], [1, 11], [1, 18], [1, 27]],
values=[10, 20, 30, 20, 11, 101, 11, 10001],
dense_shape=[2, 30])
tf.sparse.bincount(data, axis=-1)
SparseTensor(indices=tf.Tensor(
[[ 0 10]
[ 0 20]
[ 0 30]
[ 1 11]
[ 1 101]
[ 1 10001]], shape=(6, 2), dtype=int64),
values=tf.Tensor([1 2 1 2 1 1], shape=(6,), dtype=int32),
dense_shape=tf.Tensor([ 2 10002], shape=(2,), dtype=int64))
Bin-counting with defined output shape
This example takes an input (which could be a Tensor, RaggedTensor, or SparseTensor) and returns a SparseTensor where the value of (i,j) is the number of times value j appears in batch i. However, all values of j above 'maxlength' are ignored. The dense_shape of the output sparse tensor is set to 'minlength'. Note that, while the input is identical to the example above, the value '10001' in batch item 2 is dropped, and the dense shape is [2, 500] instead of [2,10002] or [2, 102].
minlength = maxlength = 500
data = np.array([[10, 20, 30, 20], [11, 101, 11, 10001]], dtype=np.int64)
tf.sparse.bincount(
data, axis=-1, minlength=minlength, maxlength=maxlength)
SparseTensor(indices=tf.Tensor(
[[ 0 10]
[ 0 20]
[ 0 30]
[ 1 11]
[ 1 101]], shape=(5, 2), dtype=int64),
values=tf.Tensor([1 2 1 2 1], shape=(5,), dtype=int64),
dense_shape=tf.Tensor([ 2 500], shape=(2,), dtype=int64))
Binary bin-counting
This example takes an input (which could be a Tensor, RaggedTensor, or SparseTensor) and returns a SparseTensor where (i,j) is 1 if the value j appears in batch i at least once and is 0 otherwise. Note that, even though some values (like 20 in batch 1 and 11 in batch 2) appear more than once, the 'values' tensor is all 1s.
data = np.array([[10, 20, 30, 20], [11, 101, 11, 10001]], dtype=np.int64)
tf.sparse.bincount(data, binary_output=True, axis=-1)
SparseTensor(indices=tf.Tensor(
[[ 0 10]
[ 0 20]
[ 0 30]
[ 1 11]
[ 1 101]
[ 1 10001]], shape=(6, 2), dtype=int64),
values=tf.Tensor([1 1 1 1 1 1], shape=(6,), dtype=int64),
dense_shape=tf.Tensor([ 2 10002], shape=(2,), dtype=int64))
Weighted bin-counting
This example takes two inputs - a values tensor and a weights tensor. These tensors must be identically shaped, and have the same row splits or indices in the case of RaggedTensors or SparseTensors. When performing a weighted count, the op will output a SparseTensor where the value of (i, j) is the sum of the values in the weight tensor's batch i in the locations where the values tensor has the value j. In this case, the output dtype is the same as the dtype of the weights tensor.
data = np.array([[10, 20, 30, 20], [11, 101, 11, 10001]], dtype=np.int64)
weights = [[2, 0.25, 15, 0.5], [2, 17, 3, 0.9]]
tf.sparse.bincount(data, weights=weights, axis=-1)
SparseTensor(indices=tf.Tensor(
[[ 0 10]
[ 0 20]
[ 0 30]
[ 1 11]
[ 1 101]
[ 1 10001]], shape=(6, 2), dtype=int64),
values=tf.Tensor([2. 0.75 15. 5. 17. 0.9], shape=(6,), dtype=float32),
dense_shape=tf.Tensor([ 2 10002], shape=(2,), dtype=int64))