While comparing two methods for simulating crossing over I was surprised to
find that the nice concise amend method below wasn't as fast or as lean as the
numerical method.
old=. i.5000 4000
new=. -old
5 4{.(old&*@:-. + new&* ) 0.7 > (5000 4000)[EMAIL PROTECTED]
0 1 2 _3
_4000 _4001 _4002 4003
8000 _8001 8002 _8003
12000 _12001 _12002 _12003
_16000 _16001 _16002 _16003
5 4{.(0.5 > 5000 [EMAIL PROTECTED])}old,:new
0 _1 2 3
_4000 4001 _4002 4003
_8000 8001 _8002 _8003
12000 _12001 _12002 _12003
_16000 16001 _16002 _16003
ts=. 6!:2,7!:[EMAIL PROTECTED]
10 ts'(old&*@:-. + new&* ) 0.9 > (5000 4000)[EMAIL PROTECTED]'
1.3008018128 704645312
10 ts'(0.9 > 5000 [EMAIL PROTECTED])}old,:new'
1.5297298050 973080320
Nevertheless I'm inclined towards the amend method for its conciseness and
because in my mind it mirrors the crossing over process better.
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
