The "row at a time" processing described in the paper was implemented in J 4 years ago in J5.04: http://www.jsoftware.com/help/release/pdt504.htm And has been in the dictionary years before that: http://www.jsoftware.com/help/dictionary/d300.htm
----- Original Message ----- From: Brian Schott <[email protected]> Date: Tuesday, June 9, 2009 11:57 Subject: Re: [Jprogramming] Inner Product -- An Old/New Problem To: Programming forum <[email protected]> > Now I see that you have shown the correct calculation in J using verb > flip and conjunction dotcj. I used power in my formulation to > emphasize the interleaving. But by doing so I may have computed all > products first. Are you suggesting that not all products in the inner > product are calculated first, but only as needed for the accumulation > (+/)? > > Also, I see that dotcj produces a dyadic hook in an explicit > definition. Very slick. > > On Tue, Jun 9, 2009 at 2:17 PM, Brian > Schott<[email protected]> wrote: > > Thanks, Roger. So does the following computation reflect the > order of > > calculation in row-at-a-time g"_1 before the f/ (that's a J /) ? > > > > 10 100 1000 <@:*"_1 i.3 4 > > +----------+---------------+---------------------+ > > |0 10 20 30|400 500 600 700|8000 9000 10000 11000| > > +----------+---------------+---------------------+ > > > > And does the following calculation reflect row-at-a-time inner > product> with a sort of interleaving? > > (+/@(2&{.),(2&}.))^:_]10 100 1000 (*"_1) i.3 4 > > 8400 9510 10620 11730 ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
