It seems I replied to a year-old thread.
Anyways, I refined your lambda into this:
tensor←{,[1 2]{,[3 4]1 3 2 4⍉⍵}⍣(1<⍴⍴⍵)(⍺∘.×⍵)}
which computes the tensor product for both vectors and matrices and is
quite readable.
Thanks,
Tobia
On Sun, Jan 24, 2016 at 10:21 PM, Louis de Forcrand
wrote:
>
I have no idea what you’re actually trying to do, but this (inelegant) lambda
works:
tensor←{ ,[1 2] { { ,[3 4] 1 3 2 4 ⍉ ⍵} ⍣ (2 < ⍴⍴⍵) ⍵ } ⍺ ∘.× ⍵}
Basically, it conditionally executes the transpose and first ravel if the
result of the outer product is of rank greater than 2.
If you d
Foad
> Date: 20 May 2015 at 14:55
> Subject: Re: [Bug-apl] tensor product
> To: Fausto Saporito
>
>
> "Why" does it give a multidimensional result? Because that's the way
> it's defined in APL. APL has generalised outer product in a way that
> works v
Wouldn't that be because ∘.× is not the tensor product operation? The
result of the ∘.OP function with arguments of rank n and m will always
yield a matrix of rank n×m.
I'd rather refer to the ∘. operator as the cartesian operator since it
applies its function on all possible permutations of the i
