CODEWITHSUNDEEP
pythonnumpy
Eric Kim
Eric Kim

Reputation: 2698

Numpy - Raise negative elements to non-integer power

import numpy as np
from scipy import stats

np.random.seed(42)
data = sorted(stats.norm.rvs(size=1000))

I want to raise data to the power of 3/4.

I tried:

np.power(data, 3/4) # doesn't work because power is not integer
np.float_power(data, 3/4) # returns nan for negative elements
scipy.linalg.fractional_matrix_power(data, 3/4) # doesn't work because data is 1D array, not matrix

How can I do this using numpy?

Upvotes: 0

Views: 1498

Answers (2)

gmds
gmds

Reputation: 19885

There is no need to call, say, np.power explicitly; by performing your computations before using sorted (or even using np.sort instead), we can just use operators:

np.sort(data ** (3 / 4))

Necessarily, this will give nan values because a root of a negative number will be complex, and your array can only handle floats.

The solution is therefore to cast it to dtype np.complex:

data = np.sort(stats.norm.rvs(size=10).astype(np.complex) ** (3 / 4))
print(data)

Output:

[-1.18840462+1.18840462j -0.10707968+0.10707968j -0.09787584+0.09787584j
 -0.05992208+0.05992208j  0.10880864+0.j          0.1484953 +0.j
  0.45161317+0.j          0.78783041+0.j          0.79189574+0.j
  0.93656538+0.j        ]

Upvotes: 1

hiew1
hiew1

Reputation: 1434

np.power(data, 3./4)

will result in a float for the power. Since 3/4 will return an int.

Upvotes: 2

Related Questions