I have an array with size (4,4) that can have values 0 and 1, so I can have 65536 different arrays. I need to produce all these arrays without repeating. I use wt_random=np.random.randint(2, size=(65536,4,4)) but I am worried they are not unique. could you please tell me this code is correct or not and what should I do to produce all possible arrays? Thank you.
4 Answers
you could use numpy.meshgrid
output = np.array(np.meshgrid(*[[0,1] for _ in range(16)])).T.reshape(-1, 4,4)
as you need to get all the possible values (number of possible values = 2 ^16 = 65536)
4 Comments
david
your suggestion is true too. Thank you.
david
is it any way to select both of proposed code as the answer?
david
is it any way to shuffle this final array? for example the first 4x4 array is all 0 and the last one is all 1s. is it any way to reorder them?
Ayoub ZAROU
No don't worry @david I'm glad I could be of help. you could do
np.random.shuffle(output)you can use itertools.product with repeat=16 to generate all patterns, then just reshape them to (4,4).
try this:
import numpy as np
from itertools import product
wt_random = np.array([np.array(p).reshape((4, 4)) for p in product((0, 1), repeat=16)])
np.random.shuffle(wt_random)
print(wt_random.shape)
print(wt_random[1234])
Output: (shows the shape is correct, and an example element)
(65536, 4, 4)
[[0 0 0 0]
[0 1 0 0]
[1 1 0 1]
[0 0 1 0]]
2 Comments
david
is it any way to shuffle this final array?
Adam.Er8
@david sure, I edited my answer docs.scipy.org/doc/numpy/reference/generated/…
You can loop through n from 1 to 65535, then have the binary equivalent mapped to your array that will be easiest and 100% truly unique and all possibilities included.
Eg for m1->0 : [ [0,0,0,0], [0,0,0,0], [0,0,0,0], [0,0,0,0] ] for m2->1 : [ [0,0,0,0], [0,0,0,0], [0,0,0,0], [0,0,0,1] ]
----------------------------------------------------
----------------------------------------------- upto n
for m65536->65535 : [ [1,1,1,1], [1,1,1,1], [1,1,1,1], [1,1,1,1] ]
2 Comments
david
this is time-consuming:( is it any way to produce all of them using the built-in function in python?
Shivam Baghla
What I would suggest you is use bin() method to convert to binary for each iteration and then can use for building array. Also if there isn't a need to produce all the arrays at once you can write a function which will return an array of particular index. Between 0-65535
If you need all possible arrays in random order, consider enumerating them in any arbitrary deterministic order and then shuffling them to randomize the order. If you don't want all arrays in memory, you could write a function to generate the array at a given position in the deterministic list, then shuffle the positions. Note that Fisher-Yates may not even need a dense representation of the list to shuffle... if you keep track of where the already shuffled entries end up you should have enough.
keras- kindly do not spam irrelevant tags (removed).