I want to generate 200 random binary arrays of size 30, but I need all of them to be unique! I use to following to generate the arrays:
import numpy as np
parents = []
for i in range(200):
parents.append(np.random.choice([0, 1], size=(30)))
parents = np.vstack(parents)
Are the arrays made in this way already unique (if so, how may I check that)? If not, how do I modify my code so I get unique arrays?
They are not unique, since if they were, then they would not be uniformly random each time.
You could perform a quick check each time you generate a new array, and make sure it is unique (using set lookups rather than array lookups for larger parent arrays):
parents = []
for i in range(200):
unique_found = False
while not unique_found:
candidate_array = np.random.choice([0, 1], size=(30))
if not any((candidate_array == x).all() for x in parents):
unique_found = True
parents.append(candidate_array)
However, since there are 1,073,741,824 unique binary arrays of length 30, the probability of getting 2 or more duplicates is:
1 - (1 - (1/1,073,741,824)) ^ (200 choose 2) = 0.0000185, or about 1 time of every 54,000.
So you might be ok with ignoring the problem.
Given that the length is large enough, the probability of generating two equal arrays is very low. Rejection sampling will be very fast:
import numpy as np
parents = set()
while len(parents) < 200:
a = tuple(np.random.choice([0, 1], size=(30)))
if a not in parents: parents.add(a)
parents = np.array([list(x) for x in parents])
And also, using set() for membership check is faster than arrays.
[0,1]and[1,0]the same arrays?