numpy.argpartition#

numpy.argpartition(a, kth, axis=-1, kind=<no value>, order=None, descending=<no value>)[source]#

Perform an indirect partition along the given axis using the algorithm specified by the kind keyword. It returns an array of indices of the same shape as a that index data along the given axis in partitioned order.

Parameters:

aarray_like: Array to sort.
kthint or sequence of ints: Element index to partition by. In the returned array, the k-th value of the array will be in the position it would be in a sorted array, all elements that are less than this element (or greater if descending is True) will be moved before it, and all elements that are greater than or equal to this element (or less than or equal if descending is True) will be moved after it. The order of all elements within each partition is undefined. If provided with a sequence of k-th it will partition all elements indexed by k-th of them into their sorted position at once.
axisint or None, optional: Axis along which to sort. The default is -1 (the last axis). If None, the flattened array is used.
kind{‘introselect’}, optional: NumPy currently offers only one selection algorithm, ‘introselect’, and this parameter provides no additional functionality. Default is None.
orderstr or list of str, optional: When a is an array with fields defined, this argument specifies which fields to compare first, second, etc. A single field can be specified as a string, and not all fields need be specified, but unspecified fields will still be used, in the order in which they come up in the dtype, to break ties.
descendingbool, optional: Sort order. If True, the array will be partitioned in descending order. If False or None, the array will be partitioned in ascending order. Values that are NaN are partitioned towards the end of the array regardless of order. Default: None.

New in version 2.6.0.

Returns:

index_arrayndarray, int: Array of indices that partition a along the specified axis. If a is one-dimensional, a[index_array] yields a partitioned a. More generally, np.take_along_axis(a, index_array, axis=axis) always yields the partitioned a, irrespective of dimensionality.

See also

partition: Describes partition algorithms used.
ndarray.partition: Inplace partition.
argsort: Full indirect sort.
take_along_axis: Apply index_array from argpartition to an array as if by calling partition.

Notes

The returned indices are not guaranteed to be sorted according to the values. Furthermore, the default selection algorithm introselect is unstable, and hence the returned indices are not guaranteed to be the earliest/latest occurrence of the element.

argpartition works for real/complex inputs with nan values, see partition for notes on the enhanced sort order and different selection algorithms.

Examples

One dimensional array:

>>> import numpy as np
>>> x = np.array([3, 4, 2, 1])
>>> x[np.argpartition(x, 3)]
array([2, 1, 3, 4]) # may vary
>>> x[np.argpartition(x, (1, 3))]
array([1, 2, 3, 4]) # may vary

>>> x = [3, 4, 2, 1]
>>> np.array(x)[np.argpartition(x, 3)]
array([2, 1, 3, 4]) # may vary

Multi-dimensional array:

>>> x = np.array([[3, 4, 2], [1, 3, 1]])
>>> index_array = np.argpartition(x, kth=1, axis=-1)
>>> # below is the same as np.partition(x, kth=1)
>>> np.take_along_axis(x, index_array, axis=-1)
array([[2, 3, 4],
       [1, 1, 3]])