Re: Tensor product question

[Jordi Gutiérrez Hermoso]

 On 15 May 2013 05:05, JFCardoso <address@hidden> wrote:
> I find Jordi's proposal a bit mysterious.

It's just broadcasting:

    http://www.gnu.org/software/octave/doc/interpreter/Broadcasting.html

> Here is an alternate version which looks more transparent, at least
> to me.

> function out = vector_tensor_product (varargin)
>   n = length (varargin);
>   dims = zeros(n,1) ;
>
>   out = varargin{1}(:);
>   dims(1) = length(out) ;
>
>   for i = 2:n
>     dims(i) = length(varargin{i})
>     out = out(:) * varargin{i}(:)' ;
>   endfor
>
>   out = reshape(out, dims) ;
> endfunction

Sure that works too. You're using broadcasting too, in a way. Ordinary
matrix multiplication of a column vector times a row vector
corresponds to broadcasting. Since Octave 3.6, a(:)*b(:)' is the same
as a(:).*b(:)'

- Jordi G. H.