How many times a number appears in a numpy array

Question

I need to find a way to count how many times each number from 0 to 9 appears in a random matrix created using np.random.randint()

import numpy as np
p = int(input("Length of matrix: "))
m = np.random.randint(0,9,(p,p))
print(m)

For example if length of matrix = 4

• [[3 4 6 5] [3 4 4 3] [4 2 4 8] [6 8 2 7]]

How many times does the number 4 appear? It should return 5.

Solution 1

You should be able to get this pretty simply:

list(m.flatten()).count(x)

Another option which is probably faster, is to use the numpy builtin count_nonzero():

np.count_nonzero(m == x)

Hooray builtin functions.

Solution 2

You can use sum function:

In [52]: m = np.random.randint(0,9,(4,4))
In [53]: m
Out[53]: 
array([[8, 8, 2, 1],
       [2, 7, 1, 2],
       [8, 6, 8, 7],
       [5, 2, 5, 2]])

In [56]: np.sum(m == 8)
Out[56]: 4

m == 8 will return a boolean array contains True for each 8 then since python evaluates the True as 1 you can sum up the array items in order to get the number of intended items.

Solution 3

If you want to get the frequency from all matrix elements, here's a simple solution using numpy.ndarray.flatten and collections.Counter:

import numpy as np
import collections

p = int(input("Length of matrix: "))
m = np.random.randint(0, 9, (p, p))
print(m)
print(collections.Counter(m.flatten()))

For example, when p=3 you'd get something like this:

[[8 4 8]
 [5 1 1]
 [1 1 1]]
Counter({1: 5, 8: 2, 4: 1, 5: 1})

Solution 4

You can flatten the matrix and then use the list count() method:

from collections import Counter
import numpy as np
p = int(input("Length of matrix: "))
m = np.random.randint(0,9,(p,p))
print(m)
flat = [item for sublist in m for item in sublist]
flat.count(4)

View more solutions

Comments

[TemporalWolf]

Counting a list is even easier than that: flat.count(x) is sufficient.

[TemporalWolf]

Also, numpy has a flatten: list(m.flatten()).count(x)

[rofls]

I'm curious which method is faster, since you still have to convert this to a list @TemporalWolf

[TemporalWolf]

Usually built-in functions are faster than custom ones. You're welcome to timeit

[Daniel]

@TemporalWolf This is incorrect, NumPy functions will be much faster than built in Python functions if the array is of a relevant size.

[TemporalWolf]

@Daniel I was including the numpy built-ins (like flatten), comparing them to rofls' custom list comprehension solution.

[Iluvatar]

This code does not count given number in array but all unique elements

[Shaig Hamzaliyev]

Changed it. Hope it works now. Thanks for pointing it out