How to convert stack of rows of binary digits into column of values in Python?

Question

Suppose I have

>>> np.array([[0,0],[0,1],[1,0],[1,1]])
array([[0, 0],
       [0, 1],
       [1, 0],
       [1, 1]])

This is a matrix, each of lines of which can be regarded as binary representation of some number, so it is

>>> np.array([[0],[1],[2],[3]])
array([[0],
       [1],
       [2],
       [3]])

How to calculate this transformation in shortest way in Python?

What have you tried so far? Please provide some samples of what you have attempted and we would be happy to help you narrow in on an optimal solution. — E. Ducateme, Jul 26 '17 at 20:45
https://stackoverflow.com/questions/15505514/binary-numpy-array-to-list-of-integers — Warren Weckesser, Jul 26 '17 at 20:55

Divakar · Accepted Answer · 2017-07-26T20:56:44.407

2

You can use matrix-multiplication with np.dot to do element-wise scaling of each column with appropriate 2-powered numbers and then sum-reduce each row, leading us to few approaches -

a.dot(2**np.arange(a.shape[1]-1,-1,-1))
a[:,::-1].dot(2**np.arange(a.shape[1]))
a.dot(1 << np.arange(a.shape[1])[::-1])

Sample run -

In [557]: a = np.array([[0,0],[0,1],[1,0],[1,1]])

In [558]: a.dot(2**np.arange(a.shape[1]-1,-1,-1))
Out[558]: array([0, 1, 2, 3])

In [559]: a[:,::-1].dot(2**np.arange(a.shape[1]))
Out[559]: array([0, 1, 2, 3])

In [566]: a.dot(1 << np.arange(a.shape[1])[::-1])
Out[566]: array([0, 1, 2, 3])

edited Jul 26 '17 at 20:56

answered Jul 26 '17 at 20:45

Divakar

218,885
19
262
358

Will read it for a week :) – Suzan Cioc Jul 26 '17 at 20:46
Can also use `np.packbits` and `np.right_shift`. As a matter of fact I think I'll add that to the answers in the non-dupe question. – Daniel F Jul 27 '17 at 06:35

How to convert stack of rows of binary digits into column of values in Python?

1 Answers1