Here's an argument against:
The dual [x] u&.v y says 'apply u, but with a different point of view.
First transform [x and] y, then apply u, then transform the result back
to the original point of view.'
The semidual x u&(a:`v) y says 'u&.v, but the point-of-view business
applies only to y'. x goes through unchanged.
What you propose puts x and y in different spaces to begin with;
presumably w and v transform them to a common point of view, but then
the result is transformed back by applying BOTH inverses. I don't see
an application for this, or an easy verbal description of it.
Henry Rich
On 11/17/2021 10:18 AM, R.E. Boss wrote:
*&.(>:`<:)/ i.2 3
|domain error
| * &.(>:`<:)/i.2 3
*&.(a:`<:)/ i.2 3
1 4 9
*&.(a:`<:)&.(>:`a:)/ i.2 3
2 6 12
so why not define
x u&.(w`v) y
as
x u&.(w`a:)&.(a:`v) y
?
R.E. Boss
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
--
This email has been checked for viruses by AVG.
https://www.avg.com
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
