tensor.mttkrp does not work with d-way tensors when d > 3

[dmdunla]

Below are the current outputs from tensor.mttkrp in pyttb and the Matlab Tensor Toolbox. For dimensions 0, 3, 4, pyttb works, but the other dimensions throw a ValueError during the reshape of the khatrirao product.

pyttb:

>>> T = ttb.tensor.from_data(np.arange(1,np.prod(shape)+1), shape)
>>> U = [];
>>> for s in shape: U.append(np.ones((s,2)))
...
>>> T.mttkrp(U,0)
array([[129600., 129600.],
       [129960., 129960.]])
>>> T.mttkrp(U,1)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "C:\Users\dmdunla\dev\github.com\pyttb\pyttb\tensor.py", line 603, in mttkrp
    Ur = np.reshape(ttb.khatrirao(U[0:self.ndims - 2], reverse=True), (szl, 1, R), order='F')
  File "<__array_function__ internals>", line 5, in reshape
  File "C:\Users\dmdunla\Anaconda3\lib\site-packages\numpy\core\fromnumeric.py", line 299, in reshape
    return _wrapfunc(a, 'reshape', newshape, order=order)
  File "C:\Users\dmdunla\Anaconda3\lib\site-packages\numpy\core\fromnumeric.py", line 58, in _wrapfunc
    return bound(*args, **kwds)
ValueError: cannot reshape array of size 48 into shape (2,1,2)
>>> T.mttkrp(U,2)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "C:\Users\dmdunla\dev\github.com\pyttb\pyttb\tensor.py", line 603, in mttkrp
    Ur = np.reshape(ttb.khatrirao(U[0:self.ndims - 2], reverse=True), (szl, 1, R), order='F')
  File "<__array_function__ internals>", line 5, in reshape
  File "C:\Users\dmdunla\Anaconda3\lib\site-packages\numpy\core\fromnumeric.py", line 299, in reshape
    return _wrapfunc(a, 'reshape', newshape, order=order)
  File "C:\Users\dmdunla\Anaconda3\lib\site-packages\numpy\core\fromnumeric.py", line 58, in _wrapfunc
    return bound(*args, **kwds)
ValueError: cannot reshape array of size 48 into shape (6,1,2)
>>> T.mttkrp(U,3)
array([[45000., 45000.],
       [48456., 48456.],
       [51912., 51912.],
       [55368., 55368.],
       [58824., 58824.]])
>>> T.mttkrp(U,4)
array([[ 7260.,  7260.],
       [21660., 21660.],
       [36060., 36060.],
       [50460., 50460.],
       [64860., 64860.],
       [79260., 79260.]])

Matlab:

>> T = tensor(1:prod(shape),shape);
>> U = {}; for i = 1:length(shape), U{i} = ones(shape(i),2); end
>> mttkrp(T,U,1)

ans =

      129600      129600
      129960      129960

>> mttkrp(T,U,2)

ans =

       86040       86040
       86520       86520
       87000       87000

>> mttkrp(T,U,3)

ans =

       63270       63270
       64350       64350
       65430       65430
       66510       66510

>> mttkrp(T,U,4)

ans =

       45000       45000
       48456       48456
       51912       51912
       55368       55368
       58824       58824

>> mttkrp(T,U,5)

ans =

        7260        7260
       21660       21660
       36060       36060
       50460       50460
       64860       64860
       79260       79260