June Kim asked:
> However, the solution isn't satisfactory to me. Is there a
> simple single step way of finding the index of the max? That
> is, the way where you don't need to scan the data twice(once
> for max, another for index).
Yes, use (i.>./) . It is supported by special code [1]:
6!:2 '(i.>./)y' NB. Optimized way
0.00146858
6!:2'y (i.>./)y' NB. Functionally identical, but
unoptimized.
0.00689049
6!:2'(i.&]>./)y' NB. Ditto
0.00731627
6!:2'(4 : ''x i. y'' >./)y' NB. Ditto
0.00697293
Relative performance in time and space:
(>A) ,. '5.2' 8!:2 (%"1 <./) 100 (6!:2 , 7!:2@:])&>
A=:'(i.>./)y';'y (i.>./)y';'(i.&]>./)y';'(4 : ''x i. y'' >./)y'
(i.>./)y 1.00 1.00
y (i.>./)y 5.04 1.11
(i.&]>./)y 5.05 1.22
(4 : 'x i. y' >./)y 5.11 3.94
-Dan
[1] http://www.jsoftware.com/help/release/idotmaxmin.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
