How to convert float array into binary?

Question

Here is my array:

array([-0.1142  ,  0.11127 ,  0.0374  ,  0.02007 , -0.05737 , -0.02058 ,
       -0.1595  , -0.1288  ,  0.1436  , -0.05212 ,  0.2437  ,  0.0046  ,
       -0.1456  , -0.09485 , -0.0788  ,  0.1755  , -0.2429  , -0.1204  ,
       -0.01064 ,  0.01154 ,  0.06058 , -0.02666 ,  0.01773 ,  0.03436 ,
       -0.1262  , -0.3428  , -0.068   , -0.10645 ,  0.0669  , -0.02094 ,
       -0.0751  ,  0.001348, -0.1737  ,  0.01146 ,  0.01648 ,  0.03613 ,
        0.03384 , -0.063   ,  0.1617  , -0.03023 , -0.258   ,  0.0385  ,
        0.0382  ,  0.1821  ,  0.2104  , -0.01604 ,  0.05945 , -0.1809  ,
        0.1847  , -0.1569  ,  0.02007 ,  0.1757  ,  0.08514 ,  0.07886 ,
       -0.00872 , -0.1108  , -0.01473 ,  0.1075  , -0.1221  ,  0.0163  ,
        0.03275 , -0.01775 ,  0.01232 , -0.0705  ], dtype=float16)

I need to convert it into binary that should recognize the decimal and negative sign then every element should output 16-bit but I don't know how please help me. I am using Jupyter Notebook to run my python program.

Answer 1

Use the struct module to interpret the half (16-bit float) as an integer, then convert that integer to a 16-character long binary string.

Per thedocumentation , ! means use the computer's system (either big-endian or little-endian) for the ordering of the two bytes, H means a unsigned short (16-bit unsigned integer), and e means a half (16-bit floating point number)

import numpy as np
import struct


arr = np.array([-0.1142  ,  0.11127 ,  0.0374  ,  0.02007 , -0.05737 , -0.02058 ,
                -0.1595  , -0.1288  ,  0.1436  , -0.05212 ,  0.2437  ,  0.0046  ,
                -0.1456  , -0.09485 , -0.0788  ,  0.1755  , -0.2429  , -0.1204  ,
                -0.01064 ,  0.01154 ,  0.06058 , -0.02666 ,  0.01773 ,  0.03436 ,
                -0.1262  , -0.3428  , -0.068   , -0.10645 ,  0.0669  , -0.02094 ,
                -0.0751  ,  0.001348, -0.1737  ,  0.01146 ,  0.01648 ,  0.03613 ,
                 0.03384 , -0.063   ,  0.1617  , -0.03023 , -0.258   ,  0.0385  ,
                 0.0382  ,  0.1821  ,  0.2104  , -0.01604 ,  0.05945 , -0.1809  ,
                 0.1847  , -0.1569  ,  0.02007 ,  0.1757  ,  0.08514 ,  0.07886 ,
                -0.00872 , -0.1108  , -0.01473 ,  0.1075  , -0.1221  ,  0.0163  ,
                 0.03275 , -0.01775 ,  0.01232 , -0.0705  ], dtype=np.float16)

def half_to_binstr(half):
    bits, = struct.unpack('!H', struct.pack('!e', half))
    return "{:016b}".format(bits)

res = np.vectorize(half_to_binstr)(arr)

Output:

array(['1010111101001111', '0010111100011111', '0010100011001010',
       '0010010100100011', '1010101101011000', '1010010101000101',
       '1011000100011011', '1011000000011111', '0011000010011000',
       '1010101010101100', '0011001111001100', '0001110010110110',
       '1011000010101001', '1010111000010010', '1010110100001011',
       '0011000110011110', '1011001111000110', '1010111110110101',
       '1010000101110011', '0010000111101001', '0010101111000001',
       '1010011011010011', '0010010010001010', '0010100001100110',
       '1011000000001010', '1011010101111100', '1010110001011010',
       '1010111011010000', '0010110001001000', '1010010101011100',
       '1010110011001110', '0001010110000101', '1011000110001111',
       '0010000111011110', '0010010000111000', '0010100010100000',
       '0010100001010101', '1010110000001000', '0011000100101101',
       '1010011110111101', '1011010000100001', '0010100011101110',
       '0010100011100100', '0011000111010100', '0011001010111100',
       '1010010000011011', '0010101110011100', '1011000111001010',
       '0011000111101001', '1011000100000101', '0010010100100011',
       '0011000110011111', '0010110101110011', '0010110100001100',
       '1010000001110111', '1010111100010111', '1010001110001011',
       '0010111011100001', '1010111111010000', '0010010000101100',
       '0010100000110001', '1010010010001011', '0010001001001111',
       '1010110010000011'], dtype='<U16')