How to select and threshold x% of values from a multi-dim NumPy array?

Question

I have a multi-dimensional NumPy array with the shape (32, 128, 128). For each entry in this array (which is of shape (128, 128), I would like to check if 80% of the values present inside it are greater than a threshold, say 0.5.

Currently, I am doing something like so:

for entry in entries: # entries: (32, 128, 128)
    raveled = np.ravel(entry) # entry: (128, 128)
    total_sum = (raveled > 0.5).sum()
    proportion = total_sum/len(raveled)

    if proportion > 0.8:
        ...

I cannot seem to figure out an efficient way to do this. Any help would be appreciated.

An entry is an array of shape (128, 128). The array is composed of floats drawn from a normal distribution with 0 mean and unit variance. — S. P
– S. P, Commented Jul 8, 2021 at 6:48
arr>0.5 creates a boolean array with the same shape. np.sum can be used to count the number of 'True/1' values. np.sum (and similar ufunc take an axis tuple to specify summing on one or more of the dimensions. — hpaulj
– hpaulj, Commented Jul 8, 2021 at 6:53

dufrmbgr · Accepted Answer · 2021-07-08 09:23:03Z

2

x =  np.random.rand(32, 128, 128)
#check 80%
np.sum(x > 0.5, axis = (1, 2)) > 0.8 * 128 * 128

x > 0.5 will return True/False boolean for all values (32 * 128 * 128). After that we are summing over 1st and 2nd axis (128 * 128) to extract the total number of True values i.e. where the conditions are met for all 32 arrays and checking whether the number is more than 80%.

edited Jul 8, 2021 at 9:23

answered Jul 8, 2021 at 8:12

dufrmbgr

4072 silver badges11 bronze badges

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

1 Comment

S. P Over a year ago

Thanks. Could you make the example a bit more comprehensive by adding comments? Might be helpful for other readers.

Collectives™ on Stack Overflow

How to select and threshold x% of values from a multi-dim NumPy array?

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related