CODEWITHSUNDEEP
pythonnumpy
Arun
Arun

Reputation: 2478

Tracking signs in numpy array

I am using Python 3.8.5 and numpy 1.18.5. I have, an example of two 5 x 3 np arrays as follows:

x
'''
array([[ 0.61671909, -3.28907688,  1.47162381],
       [-0.52380408, -0.87605603,  2.09490614],
       [-1.12907555,  0.00810023,  0.41001445],
       [-1.12316108, -0.93468271, -0.41353168],
       [-1.14203546,  0.0319224 ,  1.04925758]])
'''

y
'''
array([[-0.19258918,  0.24449053, -1.16664036],
       [-0.00495552, -1.12518086,  0.30846267],
       [-1.09738445,  1.96309078, -0.67126134],
       [-0.93575175, -0.17778996,  3.15546178],
       [ 0.73158413,  1.09391779,  0.73405069]])
'''

I want to keep track of how many of the individual (element-wise) numbers change sign from 'x' to 'y'. For example, x[0][0] changes from positive to negative in the corresponding y[0][0].

I want to compute some statistics such as: "how many of the positive numbers changed to negative from x to y", "how many of the negative numbers changed to positive from x to y", etc.

How can I achieve this?

Upvotes: 0

Views: 173

Answers (1)

Quang Hoang
Quang Hoang

Reputation: 150735

Use sign function:

np.sign(x) != np.sign(y)

If you want the coordinates, you can wrap that around np.where or np.nonzero

np.where(np.sign(x)!=np.sign(y))

If you want the total number, just sum:

np.sum(np.sign(x) != np.sign(y))
# out  6

Update for the updated question:

signs = np.sign(x) - np.sign(y)

neg2pos = np.sum( signs == -2 ) 
pos2neg = np.sum( signs == 2 )

Upvotes: 1

Related Questions