numpy group by one array and sum another

Question

I would like to take an array of numpy.bytes_ and compute a grouped sum with another array of weights. So given

x = np.array(['X', 'Y', 'X', 'Z'], dtype=np.bytes_)
w = np.array([0, 2, 1, 1])

I would like to get something like

{'X': 1, 'Y': 2, 'Z': 1}

or if numpy has named arrays that would work. The np.bytes_ array is guaranteed to be single bytes representing ASCII characters and the weights array can be integers or floats.

I've tried using some solutions here but I keep getting an exception

numpy.core._exceptions._UFuncNoLoopError: ufunc 'minimum' did not contain a loop with signature matching types (dtype('S1'), dtype('S1')) -> None

I am not familiar with numpy so I don't know what that exception means or if I should even be trying to do this with numpy. It seems like it would be pretty simple in pandas, but I thought numpy would be more efficient and I didn't want to use pandas just for this.

HYRY · Accepted Answer · 2023-04-06 05:38:43Z

3

You can use np.unique(return_inverse=True) and np.bincount(weight=...) to do this, however I don't know if this is faster than pandas.

import numpy as np
a = np.array(['X', 'Y', 'X', 'Z', 'Z', 'Y'], dtype=np.bytes_)
w = np.array([0, 2, 1, 1, 10, 100])

names, groups = np.unique(a, return_inverse=True)
counts = np.bincount(groups, weights=w)
result = dict(zip(names, counts))
print(result)

answered Apr 6, 2023 at 5:38

HYRY

97.8k28 gold badges197 silver badges192 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

ProfDFrancis · Accepted Answer · 2023-04-06 05:34:46Z

0

Use a dictionary comprehension

You can use if to exclude items with a w value of 0.

import numpy as np

x = np.array(['X', 'Y', 'X', 'Z'], dtype=np.bytes_)
w = np.array([0, 2, 1, 1])

out = {k:v for k,v in zip(x,w) if v>0}
print(out)

answered Apr 6, 2023 at 5:34

ProfDFrancis

9,4071 gold badge23 silver badges33 bronze badges

Collectives™ on Stack Overflow

numpy group by one array and sum another

2 Answers 2

Comments

Use a dictionary comprehension

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Use a dictionary comprehension

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related