Partial sorting with np.partition() in NumPy - Cheat Sheet & Quick Revision

Recall & Review

beginner

What does np.partition() do in numpy?

np.partition() rearranges elements in an array so that the element at the specified index is in its sorted position, and all smaller elements are before it, while all larger elements are after it. It does not fully sort the array.

beginner

How is np.partition() different from np.sort()?

np.sort() fully sorts the entire array, while np.partition() only ensures that the element at the given index is in the correct position, with smaller elements before and larger after, but the order within those groups is not guaranteed.

beginner

What is the main use case of np.partition()?

It is useful when you want to find the k-th smallest or largest elements quickly without sorting the entire array, such as finding medians or percentiles efficiently.

intermediate

Example: What does np.partition([3, 1, 4, 1, 5], 2) return?

It returns an array where the element at index 2 is the third smallest element, and all elements before index 2 are smaller or equal, and all after are larger or equal. One possible output is [1, 1, 3, 5, 4]. The order of elements before and after index 2 is not sorted.

intermediate

Can np.partition() be used on multi-dimensional arrays?

Yes, by specifying the axis parameter, you can partition along a specific axis of a multi-dimensional array.

What does np.partition(arr, k) guarantee about the element at index k?

A It is the k-th smallest element in the array

B The array is fully sorted

C All elements are larger than the element at index k

D The array is reversed

np.partition() places the element at index k so that it is the k-th smallest element, with smaller elements before and larger elements after, but does not fully sort the array.

Which numpy function fully sorts an array?

A np.sort()

B np.argsort()

C np.partition()

D np.argpartition()

np.sort() fully sorts the array, unlike np.partition() which only partially sorts.

If you want to find the median quickly without full sorting, which function is best?

A np.sort()

B np.cumsum()

C np.partition()

D np.argmax()

np.partition() can quickly find the median by partially sorting the array around the middle index.

What happens to elements before the partition index after using np.partition()?

A They are reversed

B They are sorted in ascending order

C They are all larger than the element at the partition index

D They are all smaller or equal to the element at the partition index

Elements before the partition index are smaller or equal to the element at that index, but not necessarily sorted.

Can np.partition() be used to find the top 3 largest elements?

A No, it only finds smallest elements

B Yes, by partitioning at the correct index

C Yes, but only for 1D arrays

D No, it only sorts fully

You can use np.partition() with negative indices or by partitioning at len(arr) - 3 to find the top 3 largest elements.

Explain how np.partition() works and when you would use it instead of full sorting.

Think about finding a specific element's position without sorting everything.

Describe the difference between np.partition() and np.sort() with an example.

Consider what happens to the array after each function.