CODEWITHSUNDEEP
pythonarraysnumpy
PiMathCLanguage
PiMathCLanguage

Reputation: 375

How to calculate a matrix raised to any power but with pretending the sign of the numbers?

Lets say the given matrix would be:

A = array([[ 5.,  6.,  4.,  1.],
           [-4., -1., -7., -4.],
           [ 2., -3.,  9.,  8.]])

and I want to raise the matrix A to the power p = 1.05.

For positive integers it is not a problem, but I want to raise the negative numbers too to the p power with pretending the negative sign (otherwise the numbers will be nan in numpy).

Here is the example of A_pow = A**p:

A_pow = array([[  5.41899193,   6.56234129,   4.28709385,   1.        ],
               [         nan,          nan,          nan,          nan],
               [  2.07052985,          nan,  10.04510857,   8.87655578]])

My approach was something like this (for the matrix B):

B = A.copy()
B[B > 0.0] = B[B > 0.0]**p
B[B < 0.0] = -np.abs(B[B < 0.0])**p

Which gives me:

B = array([[  5.41899193,   6.56234129,   4.28709385,   1.        ],
           [ -4.28709385,  -1.        ,  -7.71530221,  -4.28709385],
           [  2.07052985,  -3.16940193,  10.04510857,   8.87655578]])

Is there any other or better way to approach the same goal?

Upvotes: 2

Views: 479

Answers (1)

cs95
cs95

Reputation: 402483

Use np.abs + np.sign - 

(np.abs(A) ** 1.05) * np.sign(A)

array([[  5.41899193,   6.56234129,   4.28709385,   1.        ],
       [ -4.28709385,  -1.        ,  -7.71530221,  -4.28709385],
       [  2.07052985,  -3.16940193,  10.04510857,   8.87655578]])

np.sign retrieves the sign of each element. Then, convert the array to its absolute with np.abs, find the power, and multiply by the sign mask.

Upvotes: 4

Related Questions