From: Vassili <NOSVassili.PAMPastushenko@jku.at>
Path: news.mathworks.com!newsfeed.mathworks.com!webx
Newsgroups: comp.soft-sys.matlab
Subject: Re: What's the point of all this?
Message-ID: <eeefd44.77@webx.raydaftYaTP>
Date: Sun, 7 Nov 2004 10:09:06 -0500
References: <eeefd44.68@webx.raydaftYaTP> <eeefd44.69@webx.raydaftYaTP> <eeefd44.70@webx.raydaftYaTP> <eeefd44.74@webx.raydaftYaTP> <eeefd44.75@webx.raydaftYaTP>
Lines: 44
NNTP-Posting-Host: 140.78.123.27
MIME-Version: 1.0
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 8bit
Xref: news.mathworks.com comp.soft-sys.matlab:240081


Steve Amphlett wrote:
> The whole point of using ML in my (limited - only 14 years)
> experience is that you can prototype algorithms with the minimal
> amount of code and then, when really happy, implement them
> efficiently in a real language (i.e. one that's compiled).
> The current competition isn't really suited to ML. It's suited to
> lower level languages. Admittedly, you can write low-level code in
> ML, but why bother? GCC is free; ML isn't!
>
> IMHO, encoraging people to use ML like C or FORTRAN is wrong. I
> think there should be a serious size element to the scoring.
> - Steve

Hi Steve.
Sounds reasonable.
The best entry has about 2150 command lines! It's not a game any
more!

I do not have enough forces to participate in this game. The problem
seems to be very similar to a simple version of chess. Therefore,
minimization of the number of moves inevitably includes several
levels of analizable positions (after 1-st, second etc. moves).

I think, a smart solution should not calculate all possible versions,
but try to reduce the problem consecutively by finding and realizing
those new positions which do not influence subsequent analysis and
which allow for a shortest (optimal) trajectory to them. To find
such "points of immediate action", one can try first optimal moves of
those objects, which represent obstacles, therefore one has to
classify the optimal trajectories in terms of number of obstacles.
Additional problem is that there are many optimal trajectories within
a rectangle between start and finish, so that real skill is to write
a clever function to check whether there is a free (no obstacles)
optimal trajectory between given start and finish.

   However, already formalization of the concept "not participate in
subsequent analysis" needs some experience, which is, perhaps,
trivial for those, who did write chess-similar programs earlier, but
not to a routined user.

In any case, those who find their own working algorithm, will surely
prove that they are well above middle level programmers. I am
interested, how many people presented working wersions during dark
phase.