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 inputs. Regards, Elias On 20 May 2015 at 21:31, Fausto Saporito <[email protected]> wrote: > Hello all, > > I don't understand why the tensor product in APL (∘.×) between two > matrices, gives a multidimensional array as result. > > From linear algebra, tensor product (or kronecker product) gives a > matrix as result, > > for example > > if I2 is identity matrix > > 1 0 > 0 1 > > I2 (tensor product) I2 gives > > 1 0 0 0 > 0 1 0 0 > 0 0 1 0 > 0 0 0 1 > > instead if I use the APL operator I have : > > 1 0 > 0 1 > > 0 0 > 0 0 > > > 0 0 > 0 0 > > 1 0 > 0 1 > > the result is correct but why the shape is different ? > > Is there a way to have a behaviour similar to linear algebra result ? > > I wrote a generic tensor product function, but I'm not happy because > it uses loops and I don't like APL with loops :) but I didn't figure > out a loopless solution. > > thanks, > Fausto > >
