Check if all elements in numpy array match a number

Question

In my python testing script, I want to assert if all elements of numpy array are either very close to 1.0 or equal to 0.0. The array looks like this:

[[0.9999999991268851 1.0000000223517418 0.999999986961484 ...,
  0.9999999841675162 1.0000000074505806 0.9999999841675162]
 [0.9999999991268851 1.0000000223517418 0.999999986961484 ...,
  0.9999999841675162 1.0000000074505806 0.9999999841675162]
 [0.9999999991268851 1.0000000223517418 0.999999986961484 ...,
  0.9999999841675162 1.0000000074505806 0.9999999841675162]
 ..., 
 [1.0000000198488124 1.0000000074505806 1.000000002568413 ...,
  0.9999999888241291 0.9999999925494194 0.0]
 [1.000000011001248 0.9999999850988388 0.9999999869323801 ...,
  1.0000000186264515 0.9999999925494194 0.0]
 [1.000000011001248 0.9999999850988388 0.9999999869323801 ...,
  1.0000000186264515 0.9999999925494194 0.0]]

I thought of using numpy.allclose or numpy.array_equal, but neither makes sense here. ideally, the function should be able to be used in a testing scenario

I think you are looking at something like numpy's assert_almost_equal. See the docs for examples. — jorgeh
– jorgeh, Commented Dec 15, 2015 at 1:02

mgilson · Accepted Answer · 2015-12-15 01:07:49Z

3

You can get the 0 elements and mask them out using boolean indexing. Once that's done, np.allclose is exactly what you want:

zeros = arr == 0.0
without_zeros = arr[~zeros]
np.allclose(without_zeros, 1, ...)

answered Dec 15, 2015 at 1:07

mgilson

312k70 gold badges656 silver badges722 bronze badges

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

Comments

forkrul · Accepted Answer · 2015-12-15 01:07:49Z

0

The simplest thing I can think of would just be to iterate over every element of the array and test if it is either close to one or equal to zero:

import numpy as np

arr = np.array([[0.9999999991268851, 1.0000000223517418, 0.999999986961484],
                [1.0000000186264515, 0.9999999925494194, 0.0]])

def is_one_or_zero(arr):
  for elem in np.nditer(arr):
    if not (elem == 0 or np.isclose(1.0, elem)):
      return False
  return True

print is_one_or_zero(arr) # Should be True
arr[0, 0] = 1.01
print is_one_or_zero(arr) # Should be False

edited Dec 15, 2015 at 1:07

answered Dec 15, 2015 at 1:05

forkrul

5141 gold badge4 silver badges9 bronze badges

4 Comments

mgilson Over a year ago

I'd iterate over arr.flat instead of assuming it's a 2D array.

forkrul Over a year ago

@mgilson I changed it to use np.nditer. Is there any advantage to arr.flat versus np.nditer?

mgilson Over a year ago

I dunno :-). Probably not a huge difference between the two -- though it's been a while since I've worked with numpy heavily, so I could be wrong here...

Daniel Over a year ago

arr.flat is a fairly simple iterator that always iterates in c-contiguous order. nditer is capable of much more complex operations, see here. On a side note, python loops for numpy arrays is usually not a good idea due to the severe python performance penalty.

Collectives™ on Stack Overflow

Check if all elements in numpy array match a number

2 Answers 2

Comments

4 Comments

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

4 Comments

Related