Ants
From: Schmidt (sss_at_online.de)
Date: 07/03/04
- Next message: MikeD: "Re: MSDN Examples"
- Previous message: J French: "Re: Datatype Declaration"
- Next in thread: Schmidt: "Re: Ants"
- Reply: Schmidt: "Re: Ants"
- Messages sorted by: [ date ] [ thread ]
Date: Sat, 3 Jul 2004 13:15:01 +0200
Hi *.*
somewhat edifying for the weekend...
from june 4 to june 7 there was the seventh programming contest to compare
different languages the "non-bashing-way" (winner gets "unlimited bragging
rights" traditionally).
http://www.cis.upenn.edu/proj/plclub/contest/intro.php
The winner-statistics for languages so far (1.,2.,3. place):
1998-Cilk, OCaml, J
1999-OCaml, Haskell, Haskell
2000-OCaml, OCaml, Haskell
2001-Haskell, Dylan, Erlang
2002-OCaml, C, Python
2003-C++, C++, CDylan
2004-probably Haskell (winners will be announced in september)
Time to take part with a VB-Team next year I think. :-)
This year I heard of this contest for the first time, but unfortunately
passed the date of the contest-weekend.
The task this year seamed to be the "most interesting task ever" in history
of this contest (so the tenor in the discussion-list).
Here's the discussion list for the 2004-contest, just to get an impression
of the "flow".
http://lists.seas.upenn.edu/pipermail/icfp04-discuss/2004-June/date.html
The task is (in simple words):
Use your language of choice, to generate a textfile with max. 10000 rows,
every row representing a single state of an "ant-brain" (the description of
a finite state machine).
The ant-files (ant-brains) will then be compared (one vs one) by the jury
inside a simulator with randomly generated "ant-worlds" - winner is the
ant-colony, wich has transported the most food into its ant-hill after
100000 rounds of simulation.
The task descrption http://www.cis.upenn.edu/proj/plclub/contest/ants.html,
clearly defines the possible states of the "ant-brain", the simulator, the
worlds and all other things one has to take into account, to generate own
ants.
Everyone who wants to make good ants has to write a fast, "jury-compatible"
simulator first.
Nearly everyone stated (and so do I), that implementing this one was most
pleasant and pure fun.
One can verify, that his simulator is ok, comparing the dump output after
10.000 rounds (for a given ant and a given tiny world) with the given
dump-output of the ICFP.
I had mine up and running after 7 hours and needed additional 4 hours for
bugfixing and performance-tuning.
If this is done one can write own ants or let generate ants by genetic
algos.
(The best "human-brain" ants seem to perform better than those, grown by
genetic algos so far)
Have fun!
Olaf
I will post a link to my sources tomorrow.
- Next message: MikeD: "Re: MSDN Examples"
- Previous message: J French: "Re: Datatype Declaration"
- Next in thread: Schmidt: "Re: Ants"
- Reply: Schmidt: "Re: Ants"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
