Forget it,  ws   is wrong.  If you can figure out a function of a vector of 
weights  w  that will ensure  </ ? (i,j){w  is exactly as likely as  %/(i,j){w  
then you can replace  (+ +/)  with that function.  But I suppose that begs the 
question.

-Dan

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

Reply via email to