> Van: [email protected] [mailto:programming-
> [email protected]] Namens Raul Miller
(...)
> 
> Here is a model for a "general oblique":
> 
> oblq=:1 :0
>   assert. (=<.)rnk2=:-:#$y
>   s1=: rnk2 {. $y
>   s2=: rnk2 }. $y
>   inds=: i.<:s1+s2
>   sel=: (s1{.inds) +/&, s2 {. inds
>   ($inds) $ (~.,sel) /:~ (,sel) u/.,y
> )
> 
(...)
> 
> [2] Should J be changed to support this "more
> general" approach to convolution?  If so, what
> motivation do we need and what changes would
> this imply?
> 

No. 

   <oblq i.2 3 4
|assertion failure
|   (=<.)rnk2=:-:#$y


obl=: 1 : '(+/"1)@(#: [:i. */)@$ u/. ,'

   <obl i.2 3 4
+-+------+-----------+--------------+-------------+--------+--+
|0|1 4 12|2 5 8 13 16|3 6 9 14 17 20|7 10 15 18 21|11 19 22|23|
+-+------+-----------+--------------+-------------+--------+--+


R.E. Boss

----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to