top_k#

ivy.top_k(x, k, /, *, axis=-1, largest=True, sorted=True, out=None)[source]#

Return the k largest elements of the given input array along a given axis.

Parameters:

x (Union[Array, NativeArray]) – The array to compute top_k for.
k (int) – Number of top elements to retun must not exceed the array size.
axis (int) – The axis along which we must return the top elements default value is 1. (default: -1)
largest (bool) – If largest is set to False we return k smallest elements of the array. (default: True)
sorted (bool) – If sorted is set to True we return the elements in sorted order. (default: True)
out (Optional[tuple]) – Optional output tuple, for writing the result to. Must have two arrays inside, (default: None) with a shape that the returned tuple broadcast to.

Return type:

Tuple[Array, NativeArray]

Returns:

ret – A named tuple with values and indices of top k elements.

Examples

With ivy.Array input:

>>> x = ivy.array([2., 1., -3., 5., 9., 0., -4])
>>> y = ivy.top_k(x, 2)
>>> print(y)
top_k(values=ivy.array([9., 5.]), indices=ivy.array([4, 3]))

>>> x = ivy.array([[-2., 3., 4., 0.], [-8., 0., -1., 2.]])
>>> y = ivy.top_k(x, 2, axis=1, largest=False)
>>> print(y)
top_k(values=ivy.array([[-2.,  0.],[-8., -1.]]),
...   indices=ivy.array([[0, 3],[0, 2]]))

With ivy.NativeArray input:

>>> x = ivy.native_array([2., 1., -3., 5., 9., 0., -4])
>>> y = ivy.top_k(x, 3)
>>> print(y)
top_k(values=ivy.array([9., 5., 2.]), indices=ivy.array([4, 3, 0]))

With ivy.Container input:

>>> x = ivy.Container(a=ivy.array([-1, 2, -4]), b=ivy.array([4., 5., 0.]))
>>> y = ivy.top_k(2)
>>> print(y)
{
    a: [
        values = ivy.array([ 2, -1]),
        indices = ivy.array([1, 0])
    ],
    b: [
        values = ivy.array([5., 4.]),
        indices = ivy.array([1, 0])
    ]
}

Array.top_k(self, k, /, *, axis=-1, largest=True, sorted=True, out=None)#

ivy.Array instance method variant of ivy.top_k. This method simply wraps the function, and so the docstring for ivy.top_k also applies to this method with minimal changes.

Parameters:

self (Array) – The array to compute top_k for.
k (int) – Number of top elements to retun must not exceed the array size.
axis (int) – The axis along which we must return the top elements default value is 1. (default: -1)
largest (bool) – If largest is set to False we return k smallest elements of the array. (default: True)
sorted (bool) – If sorted is set to True we return the elements in sorted order. (default: True)
out (Optional[tuple]) – Optional output tuple, for writing the result to. Must have two arrays, (default: None) with a shape that the returned tuple broadcast to.

Return type:

Tuple[Array, NativeArray]

Returns:

ret – A named tuple with values and indices of top k elements.

Examples

With ivy.Array input:

>>> x = ivy.array([2., 1., -3., 5., 9., 0., -4])
>>> y = x.top_k(2)
>>> print(y)
top_k(values=ivy.array([9., 5.]), indices=ivy.array([4, 3]))

Container.top_k(self, k, /, *, axis=-1, largest=True, sorted=True, key_chains=None, to_apply=True, prune_unapplied=False, map_sequences=False, out=None)#

ivy.Container instance method variant of ivy.top_k. This method simply wraps the function, and so the docstring for ivy.top_k also applies to this method with minimal changes.

Parameters:

self (Container) – The container to compute top_k for.
k (int) – Number of top elements to retun must not exceed the array size.
axis (int) – The axis along which we must return the top elements default value is 1. (default: -1)
largest (bool) – If largest is set to False we return k smallest elements of the array. (default: True)
sorted (bool) – If sorted is set to True we return the elements in sorted order. (default: True)
key_chains (Optional[Union[List[str], Dict[str, str]]]) – The key-chains to apply or not apply the method to. Default is None. (default: None)
to_apply (bool) – If True, the method will be applied to key_chains, otherwise key_chains (default: True) will be skipped. Default is True.
prune_unapplied (bool) – Whether to prune key_chains for which the function was not applied. (default: False) Default is False.
map_sequences (bool) – Whether to also map method to sequences (lists, tuples). (default: False) Default is False
out (Optional[Tuple[Container, Container]]) – Optional output tuple, for writing the result to. Must have two Container, (default: None) with a shape that the returned tuple broadcast to.

Return type:

Tuple[Container, Container]

Returns:

ret – a container with indices and values.

Examples

With ivy.Container input:

>>> x = ivy.Container(a=ivy.array([-1, 2, -4]), b=ivy.array([4., 5., 0.]))
>>> y = x.top_k(2)
>>> print(y)
{
    a: [
        values = ivy.array([ 2, -1]),
        indices = ivy.array([1, 0])
    ],
    b: [
        values = ivy.array([5., 4.]),
        indices = ivy.array([1, 0])
    ]
}