On Tue, Oct 21, 2014 at 2:54 PM, Mike Day <[email protected]>
wrote:
> No takers to the previous posting (below)?
Sorry, I'm busy on another project at the moment, and do not feel
motivated to take the time to think through this issue.
That said, I guess I would use
([:(*/"1) 1 + [: {:"2 (__&q:)"1)
as my starting point, and I'd probably want to apply it to reasonably large
blocks of numbers (maybe a million at a time), and I'd probably duplicate
my adjacent blocks or something to deal with the whole "scanning a range of
numbers which would necessarily span blocks" thing.
In other words I'd run my brute force max reduce sum scans on
something modeled after:
2 ,/\ i. 4 4
0 1 2 3 4 5 6 7
4 5 6 7 8 9 10 11
8 9 10 11 12 13 14 15
If that was still "too slow", I'd start thinking about patterns in the
data with an eye towards analytically eliminating some of the brute
aspects.
Thanks,
--
Raul
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
