I have two numpy arrays with shapes (5741, 20000) and (5741, 11) respectively. How can I concatenate them into one array of shape (5741, 2), i.e. each row of the newly created array should contain the row of the first array and the row of the second array?!
Ex:
A = [(1, 1),
(2, 2),
(3, 3)]
B = [(A),
(B),
(C)]
new_array = [((1, 1), (A)),
((2, 2), (B)),
((3, 3), (C))]
Use numpy.hstack():
>>> A=np.array([[1,1],[2,2],[3,3]])
>>> B=np.array([[100],[200],[300]])
>>> np.hstack((A,B))
array([[ 1, 1, 100],
[ 2, 2, 200],
[ 3, 3, 300]])
In [65]: A = np.arange(6).reshape(3,2)
In [66]: B = np.arange(3).reshape(3,1)
hstack or concatenate on last dimension produces a (n,3) array
In [67]: np.concatenate((A,B),axis=1)
Out[67]:
array([[0, 1, 0],
[2, 3, 1],
[4, 5, 2]])
To create an array that contains subarrays of differing length is trickier. A good starting point is an object array of the desired shape:
In [68]: C = np.empty((3,2),dtype=object)
In [69]: C[:,0] = A
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-69-928e3f2975cf> in <module>()
----> 1 C[:,0] = A
ValueError: could not broadcast input array from shape (3,2) into shape (3)
But even with that copying arrays into the object slots can be tricky. Copying lists to the slots is easier:
In [70]: C[:,0] = A.tolist()
In [71]: C[:,1] = B.tolist()
In [72]: C
Out[72]:
array([[list([0, 1]), list([0])],
[list([2, 3]), list([1])],
[list([4, 5]), list([2])]], dtype=object)
With a bit more fiddling I could make that into an array of arrays rather than list. Or maybe array of tuples?
Do you really understand what that (n,2) array will be involve?
[((1, 1), (A)),
((2, 2), (B)),
((3, 3), (C))]
Another approach is a structured array:
In [74]: D = np.empty(3, dtype=[('x',int,2), ('y',int)])
In [75]: D['x']=A
In [76]: D['y']=B
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-76-fbab1c580883> in <module>()
----> 1 D['y']=B
ValueError: could not broadcast input array from shape (3,1) into shape (3)
In [77]: D['y']=B.flat
In [78]: D
Out[78]:
array([([0, 1], 0), ([2, 3], 1), ([4, 5], 2)],
dtype=[('x', '<i8', (2,)), ('y', '<i8')])
Another way of writing A and B to C:
In [81]: C[:,1]=B.ravel()
In [83]: for i in range(3): C[i,0]=A[i]
In [84]: C
Out[84]:
array([[array([0, 1]), 0],
[array([2, 3]), 1],
[array([4, 5]), 2]], dtype=object)
or writing tuples:
In [85]: for i in range(3): C[i,0]=tuple(A[i])
In [87]: for i in range(3): C[i,1]=tuple(B[i])
In [88]: C
Out[88]:
array([[(0, 1), (0,)],
[(2, 3), (1,)],
[(4, 5), (2,)]], dtype=object)
