I also have 2.4 Ghz pentium

For 100 all 1's its 0.09 seconds
For 100 all 0's also 0.09 seconds
This a not very meaningful because the algorithm spots matrices with
zero determinats early.

More meaningful
For 100 all random its 0.494 seconds (including the time to create
the random numbers. which is about 0.1 seconds)

The algorithm is one I have designed myself and despite extensive
research I have not seen it anywhere on the web. The idea is similar
to the algorithm fred is using. I have been having problems with
rounding errors though so I might try an algorithm like LU
decomposition.

I suppose I better use this to get some better results!

:)

Yes, nobody can't find if your solutions comes from other's code.
But you will have much more fun using the same set of rules as your
opponents do.
Obviously, you can download even matrices, and people won't be able
to catch you.

--- In larscontest@yahoogroups.com, "pseudosphinx"
<pseudosphinx@y...> wrote:
>
> about the code,i can't understand why you say don't use other ppl
> code if you want only solution and not code to it.
> (the ideea is that its lame to CTRL+C CTRL+V somebody's else's
code
> but i can't understand how could you see if my solution comes from
> that persons code).

----- Original Message -----

From: Jean-Charles Meyrignac

To: larscontest@yahoogroups.com

Sent: Thursday, March 03, 2005 2:35 PM

Subject: [larscontest] Benchmarks

I know it's a little bit too early, but I would like to know how fast you
can compute determinants for part A of the contest.
Of course, you don't have to disclose your technique, but can you post the
time it takes to compute the determinant of a 59x59 matrix filled with 0 and
a 59x59 matrix filled with 1 ?
How much time does it take with Mathematica ?

At this moment, I've a very slow routine, and it's really slow even on 20x20
matrices !

JC

I know it's a little bit too early, but I would like to know how fast you
can compute determinants for part A of the contest.
Of course, you don't have to disclose your technique, but can you post the
time it takes to compute the determinant of a 59x59 matrix filled with 0 and
a 59x59 matrix filled with 1 ?
How much time does it take with Mathematica ?

At this moment, I've a very slow routine, and it's really slow even on 20x20
matrices !

JC

I think that posting already found results won't change the final contest
results.
For example, how can you check if a matrix is already known ?
BTW, it could be a good way to encourage new participants, since writing a
routine to compute the determinant is hard.

JC

>
> I think the point in the rule really boils down to
> good sportsmanship.  It's true that it is impossible
> to tell if a solution was copied (either the solution
> or the code that generated it), but what fruit is
> there in submitting answers that weren't from your own
> ideas or implementations?
>
> -timmac
>
>

I think the point in the rule really boils down to
good sportsmanship.  It's true that it is impossible
to tell if a solution was copied (either the solution
or the code that generated it), but what fruit is
there in submitting answers that weren't from your own
ideas or implementations?

-timmac



--- pseudosphinx <pseudosphinx@...> wrote:

>
> about the code,i can't understand why you say don't
> use other ppl
> code if you want only solution and not code to it.
> (the ideea is that its lame to CTRL+C CTRL+V
> somebody's else's code
> but i can't understand how could you see if my
> solution comes from
> that persons code).
>
>
>
>
>


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

about the code,i can't understand why you say don't use other ppl
code if you want only solution and not code to it.
(the ideea is that its lame to CTRL+C CTRL+V somebody's else's code
but i can't understand how could you see if my solution comes from
that persons code).

I thought about this, and I don't think I'm going to do it.  I think that part
of the game is figuring out where you need to improve.

-Lars


On 3/2/2005 at 5:16 AM Radu Grigore wrote:

>Is it possible to see individual _normalized_ scores? I think it would
>help competitors to know on which values of N to concentrate.
>

You can use standard libaries that are included with standard distributions of
whatever language you are using.  Other than that though, you should write your
own code.

-Lars


On 3/2/2005 at 3:42 PM Nagi Nahas wrote:

>Not download someone else's code ? So , using linear
>algebra algorithm libraries is forbidden ?

--- Lars Backstrom <contest@...>
wrote:

> [...] You just shouldn't
> download the matrices straight off of another
> webpage, or download someone else's code.
>
> -Lars
>
>

Not download someone else's code ? So , using linear
algebra algorithm libraries is forbidden ?





__________________________________
Celebrate Yahoo!'s 10th Birthday!
Yahoo! Netrospective: 100 Moments of the Web
http://birthday.yahoo.com/netrospective/

It would be nice to see also rankings for all people - like 1st,
2nd, 23rd place.

Is it possible to see individual _normalized_ scores? I think it would
help competitors to know on which values of N to concentrate.

While using matrices you find online or elsewhere is not allowed, there is a
fair amount of literature pertaining to the problem, which provides some good
insights.  I've been asked if it is ok to use these resources and the answer is
yes.  If you learn about various techniques for finding good matrices, you are
welcome to use them.  You just shouldn't download the matrices straight off of
another webpage, or download someone else's code.

-Lars

This is fixed.


On 3/1/2005 at 8:02 AM tomrokicki wrote:

>--- In larscontest@yahoogroups.com, "olexiy_oryeshko"
><olexiy_oryeshko@y...> wrote:
>>
>> As I see now in the part B:
>> Tim MacDonald 0.0097 0 2005-02-28 14:38:28
>>
>> So, Tim MacDonald has 0 submissions, which is probably wrong.
>
>Right, and in the past few hours I've lost a full point in part B, and
>the only submissions in that time were the 03/01 submissions neither
>of whom have anywhere near even a single point.  So something's broken.
>
>
>
>
>
>
>Yahoo! Groups Links
>
>
>

see the bottom of the problem page.  You start with a 1 for part b, since its
trivial to achieve that.  Submissions only shows submissions that improve your
score.
-Lars


On 3/1/2005 at 7:22 AM olexiy_oryeshko wrote:

>As I see now in the part B:
>Tim MacDonald 0.0097 0 2005-02-28 14:38:28
>
>So, Tim MacDonald has 0 submissions, which is probably wrong.
>
>
>
>
>
>
>Yahoo! Groups Links
>
>
>

--- In larscontest@yahoogroups.com, "olexiy_oryeshko"
<olexiy_oryeshko@y...> wrote:
>
> As I see now in the part B:
> Tim MacDonald 0.0097 0 2005-02-28 14:38:28
>
> So, Tim MacDonald has 0 submissions, which is probably wrong.

Right, and in the past few hours I've lost a full point in part B, and
the only submissions in that time were the 03/01 submissions neither
of whom have anywhere near even a single point.  So something's broken.

As I see now in the part B:
Tim MacDonald 0.0097 0 2005-02-28 14:38:28

So, Tim MacDonald has 0 submissions, which is probably wrong.

1) You'll get a 0
2) Please only submit matrices you've found using on your own.  I think that
because of the sparsity condition, the winners will end up doing better than
someone who submitted only matrices they found from others anyway.

-Lars


On 2/28/2005 at 8:30 PM Jean-Charles Meyrignac wrote:

>Two questions:
>
>1) part A: what happens if I submit a matrix with only 0s ?
>Will I get an infinite score, or a divide by zero ?
>
>2) In the rules, it is written:
> You're welcome to use any resources you desire to, but you may not submit
>answers found by others.
>
>Can I submit the matrices found by mathematicians after the 100+ years of
>this problem ?
>
>JC
>

Two questions:

1) part A: what happens if I submit a matrix with only 0s ?
Will I get an infinite score, or a divide by zero ?

2) In the rules, it is written:
  You're welcome to use any resources you desire to, but you may not submit
answers found by others.

Can I submit the matrices found by mathematicians after the 100+ years of
this problem ?

JC


>
> Sorry to spring this on you guys after the start, but I guess I didn't
> really do all my homework, and part A is pretty well researched, and
> there are lots of good matrices available if you search oneline, which
> will likely be very difficult to improve on.
>
> After some thought, I've decided to tweak the scoring in such a way
> that the problem is similar, but people can not just download matrices.
>
> Before, for part A, your score was the determinant of the matrix you
> submitted.  I'm going to change this to be the determinant of the
> matrix, divided by the square of the number of 1's in the matrix.
> Hence, matrices that are more sparse will receive higher scores.
>
> Hopefully, this won't effect anyone very much, and will make the
> results more interesting.  Again, I apoligize for changing the contest
> in the middle, but I don't think anyone would stand much chance
> against 100 years of serious research!
>
> I'll fix the scoring system and rescore the results I hava ASAP.
>
> -Lars

--- In larscontest@yahoogroups.com, "lbackstrom" <contest@w...> wrote:
> Before, for part A, your score was the determinant of the matrix you
> submitted.  I'm going to change this to be the determinant of the
> matrix, divided by the square of the number of 1's in the matrix.

The short introduction of the problems still says that for part A we
should maximize the value of the determinant. Maybe the "divided
by..." part should be added here too.

regards,
  radu

Sorry to spring this on you guys after the start, but I guess I didn't
really do all my homework, and part A is pretty well researched, and
there are lots of good matrices available if you search oneline, which
will likely be very difficult to improve on.

After some thought, I've decided to tweak the scoring in such a way
that the problem is similar, but people can not just download matrices.

Before, for part A, your score was the determinant of the matrix you
submitted.  I'm going to change this to be the determinant of the
matrix, divided by the square of the number of 1's in the matrix.
Hence, matrices that are more sparse will receive higher scores.

Hopefully, this won't effect anyone very much, and will make the
results more interesting.  Again, I apoligize for changing the contest
in the middle, but I don't think anyone would stand much chance
against 100 years of serious research!

I'll fix the scoring system and rescore the results I hava ASAP.

-Lars

no, you were right.  When I fixed the last bug, I broke that.  I clearly have no
understanding of the way php's strpos function works.

I think it should be fixed now.

-Lars


On 2/27/2005 at 1:41 PM Borisas Bursteinas wrote:

>there is no bug. I just did not check properly how the entries for
>part B are calculated.
>
>
>
>
>
>
>Yahoo! Groups Links
>
>
>

there is no bug. I just did not check properly how the entries for
part B are calculated.

In part B it does not detect the best value.

--- In larscontest@yahoogroups.com, "Lars Backstrom" <contest@w...> wrote:
> Can you try to submit it again.  I think I tracked down the bug.  If
it still doesn't work, email me the case, so I can figure out what's
going on.
>
> -Lars
>
>
> On 2/27/2005 at 10:43 AM Borisas Bursteinas wrote:
>
> >Lars, could you check why my entry for N=10 shows wrong best value. It
> >calculates it correctly when I submit, but do not update the best
> >values with it.
> >
> >
> >
> >
> >
> >
> >Yahoo! Groups Links
> >
> >
> >

It works now, :) thenks.

Borka

Can you try to submit it again.  I think I tracked down the bug.  If it still
doesn't work, email me the case, so I can figure out what's going on.

-Lars


On 2/27/2005 at 10:43 AM Borisas Bursteinas wrote:

>Lars, could you check why my entry for N=10 shows wrong best value. It
>calculates it correctly when I submit, but do not update the best
>values with it.
>
>
>
>
>
>
>Yahoo! Groups Links
>
>
>

Lars, could you check why my entry for N=10 shows wrong best value. It
calculates it correctly when I submit, but do not update the best
values with it.

The contest has started!
You can register and submit your solutions at
www.worlddesigncenter.com/contest/

If you want to buy some antique reproduction furniture, thats there
too, up one directory. :)

I'm a bit new to all this web programming stuff, so if you manage to
break anything, send me an email at contest@..., and
I'll do my best to fix it promptly.

Good luck, and happy coding.

-Lars