CODEWITHSUNDEEP
pythonnumpyslice
Captain Trojan
Captain Trojan

Reputation: 2921

Slicing an [m, n] array by a variable [m] array in numpy

Let there be a numpy array of shape [M], dtype int32 and (random) values in range [0, N), e.g.:

M = 8
N = 5
a = np.random.randint(0, N, [M])  # a = [1, 1, 2, 4, 0, 1, 1, 3]

From this array I need to create a matrix m of shape [M, N], dtype int32 and values 0 or 1, where m[i,j] = 0 if j < a[i], otherwise 1. Following the example:

m = some_magic(a)     # m = [[0, 1, 1, 1, 1],
                      #      [0, 1, 1, 1, 1],
                      #      [0, 0, 1, 1, 1],
                      #      [0, 0, 0, 0, 1],
                      #      [1, 1, 1, 1, 1],
                      #      [0, 1, 1, 1, 1],
                      #      [0, 1, 1, 1, 1],
                      #      [0, 0, 0, 1, 1]]

My dysfunctional version of some_magic starts with initializing the matrix to zeros (using np.zeros), and then proceeding to set the appropriate members to 1.

m = np.zeros([M, N])

This next part though I cannot properly figure out. Accessing single members, for example, every second member, or a fixed slice, is easy, and achievable by

m[np.arange(M), C1:C2]

where C1 and C2 are integer constants,

m[np.arange(M), a:]

which, as far as I've thought, should yield the correct result, fails with the error being

Only integer scalar arrays can be converted to a scalar index.

Can you please point me to the right direction? Thank you very much.

Upvotes: 1

Views: 209

Answers (2)

user3483203
user3483203

Reputation: 51155

Here's a solution using broadcasting.

(a[:, None] <= np.arange(N)).view('i1')
# np.less_equal.outer(a, np.arange(N)).view('i1')


array([[0, 1, 1, 1, 1],
       [0, 1, 1, 1, 1],
       [0, 0, 1, 1, 1],
       [0, 0, 0, 0, 1],
       [1, 1, 1, 1, 1],
       [0, 1, 1, 1, 1],
       [0, 1, 1, 1, 1],
       [0, 0, 0, 1, 1]], dtype=int8)

Upvotes: 1

peer
peer

Reputation: 4699

I'm not sure if slicing like that is possible. I suggest you create indices instead and then operate on those:

M = 8 
N = 5 
#a = np.random.randint(0, N, [M])  
a = np.array([1, 1, 2, 4, 0, 1, 1, 3])

from0toN = np.expand_dims(np.arange(N),0) # [[0,1,2,3,4]]

m = np.repeat(from0toN, M, axis=0)
#array([[0, 1, 2, 3, 4],
#             ...,
#       [0, 1, 2, 3, 4]])

boolean = m >= np.expand_dims(a,1)

onesAndZeroes = boolean.astype(int)

"""
array([[0, 1, 1, 1, 1],
       [0, 1, 1, 1, 1],
       [0, 0, 1, 1, 1],
       [0, 0, 0, 0, 1],
       [1, 1, 1, 1, 1],
       [0, 1, 1, 1, 1],
       [0, 1, 1, 1, 1],
       [0, 0, 0, 1, 1]])
"""

Upvotes: 0

Related Questions