From: "Yi Cao" <>
Newsgroups: comp.soft-sys.matlab
Subject: Re: Hole, Please correct it.
Message-ID: <eeb4a65.48@WebX.raydaftYaTP>
Date: Tue, 12 Nov 2002 08:50:31 -0500
References: <eeb4a65.41@WebX.raydaftYaTP> <eeb4a65.42@WebX.raydaftYaTP> <eeb4a65.43@WebX.raydaftYaTP> <eeb4a65.46@WebX.raydaftYaTP> <eeb4a65.47@WebX.raydaftYaTP>
Lines: 38
MIME-Version: 1.0
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 8bit
Xref: comp.soft-sys.matlab:142629

Heinrich Acker wrote:
> It would be nice if the MATLAB contest team
> could provide an explanation of 'the hole'
> (after fixing it) that is clear not only
> for the pros. I didn't get it yet.
> 1) How is it possible that all the leading
> entries couldn't finish the test suit at
> home?
> 2) How is it possible that an entry stores
> information, but a tweak of that entry
> does not benefit?
> Heinrich

1) I think you meant the code originally contributed by Lucio
Andrade. There was a bug there. Now it has been fixed by Yuval. See
I guess why this bug can pass contest check but cannot run on our
testsuit is due the difference between these two testsuits.
2) The code, which identifing the hole, stores current results if the
global variable is empty, or just returns the previous results if the
global variable exists and the protein is matched. Therefore, it is
understandable why a tweak may have no effect on improvement because
it depends when do you submit the tweak, or exactly whom does your
tweak follows. You can devise this hole/code by a double-submision
with new global variable names as Roger did.

My code originally comes from the idea to check if the testsuit has
identical proteins. If so, why we need repeatedly calculating.
According to previous contest policy, they reboot PC before starting
very new entry. If so, this code is OK. Every entry needs creating
new global variables once. Surprisingly I found they changed the
policy and the code has made chaos. I am sorry for that.