Create 3D array from a 2D array by replicating/repeating along the first axis

Question

Suppose I have a n × m array, i.e.:

array([[ 1.,  2.,  3.],
       [ 4.,  5.,  6.],
       [ 7.,  8.,  9.]])

And I what to generate a 3D array k × n × m, where all the arrays in the new axis are equal, i.e.: the same array but now 3 × 3 × 3.

array([[ 1.,  2.,  3.],
       [ 4.,  5.,  6.],
       [ 7.,  8.,  9.]],

      [[ 1.,  2.,  3.],
       [ 4.,  5.,  6.],
       [ 7.,  8.,  9.]],

      [[ 1.,  2.,  3.],
       [ 4.,  5.,  6.],
       [ 7.,  8.,  9.]]])

How can I get it?

Answer 1

Introduce a new axis at the start with None/np.newaxis and replicate along it with np.repeat. This should work for extending any n dim array to n+1 dim array. The implementation would be -

np.repeat(arr[None,...],k,axis=0)

Sample run -

In [143]: arr
Out[143]: 
array([[ 1.,  2.,  3.],
       [ 4.,  5.,  6.],
       [ 7.,  8.,  9.]])

In [144]: np.repeat(arr[None,...],3,axis=0)
Out[144]: 
array([[[ 1.,  2.,  3.],
        [ 4.,  5.,  6.],
        [ 7.,  8.,  9.]],

       [[ 1.,  2.,  3.],
        [ 4.,  5.,  6.],
        [ 7.,  8.,  9.]],

       [[ 1.,  2.,  3.],
        [ 4.,  5.,  6.],
        [ 7.,  8.,  9.]]])

View-output for memory-efficiency

We can also generate a 3D view and achieve virtually free runtime with np.broadcast_to. More info - here. Hence, simply do -

np.broadcast_to(arr,(3,)+arr.shape) # repeat 3 times