Greetings,
At the moment I am working on finding an optimal cover
for L(21,6,4,4) ie There are 21 balls in a barrel, 6 are
drawn and I need a set of pick 4's that will deliver at least
one hit of 4 no matter which of the six numbers are drawn.
I have developed a program that can solve this in about
900 lines (ie 900 pick fours). In the initial stages it tries to
avoid pick fours that have common internal combinations
of three with other picks. this method is exhausted
the program uses a "greedy" method to complete the cover.
Simmulated Annealing and or Tabu Search are said to be
the preferred methods for these type of problems. I have
done extensive web searches but cannot find a site which
explains these techniques in a manner that I can understand.
Anyone out there with sufficient understanding of
SA and Tabu who could apply the techniques to my problem?
A detailed algorithm would be sufficient, I will do the coding
in Vb6 or C
Regards Michael Harrington.

Mon, 4 Sep 2006 12:25:07 +1000, "Michael Harrington"
<mikharr@bigpond.comwrote:


>
>Anyone out there with sufficient understanding of
>SA and Tabu who could apply the techniques to my problem?
>A detailed algorithm would be sufficient, I will do the coding
>in Vb6 or C

Consulting? Market price is about $300 per hour. A week would be
probably K.

A.L.

Michael Harrington wrote:

Greetings,

At the moment I am working on finding an optimal cover
for L(21,6,4,4) ie There are 21 balls in a barrel, 6 are
drawn and I need a set of pick 4's that will deliver at least
one hit of 4 no matter which of the six numbers are drawn.

I have developed a program that can solve this in about
900 lines (ie 900 pick fours). In the initial stages it tries to
avoid pick fours that have common internal combinations
of three with other picks. this method is exhausted
the program uses a "greedy" method to complete the cover.

Simmulated Annealing and or Tabu Search are said to be
the preferred methods for these type of problems. I have
done extensive web searches but cannot find a site which
explains these techniques in a manner that I can understand.

The wikipedia entries are not very detailed but easy to understand. It may
get you started:

Anyone out there with sufficient understanding of
SA and Tabu who could apply the techniques to my problem?
A detailed algorithm would be sufficient, I will do the coding
in Vb6 or C

Regards Michael Harrington.

"Alex R. Mosteo" <devnull@mailinator.comwrote in message
news:4m2p2nF484mkU3@individual.net
Michael Harrington wrote:
>
>Greetings,
>>
>At the moment I am working on finding an optimal cover
>for L(21,6,4,4) ie There are 21 balls in a barrel, 6 are
>drawn and I need a set of pick 4's that will deliver at least
>one hit of 4 no matter which of the six numbers are drawn.
>>
>I have developed a program that can solve this in about
>900 lines (ie 900 pick fours). In the initial stages it tries to
>avoid pick fours that have common internal combinations
>of three with other picks. this method is exhausted
>the program uses a "greedy" method to complete the cover.
>>
>Simmulated Annealing and or Tabu Search are said to be
>the preferred methods for these type of problems. I have
>done extensive web searches but cannot find a site which
>explains these techniques in a manner that I can understand.
>
The wikipedia entries are not very detailed but easy to understand. It may
get you started:


>
>Anyone out there with sufficient understanding of
>SA and Tabu who could apply the techniques to my problem?
>A detailed algorithm would be sufficient, I will do the coding
>in Vb6 or C
>>
>Regards Michael Harrington.
>

Thanks Alex, that is a promising link.
My first major hurdle is understanding the
concept of "neighboring solution" My intitial
solution is a list of 900 pick 4's. What would
be a neighboring solution to this be? Would
it be the random tweaking of just one of the
900 elements, provided this "tweak" still
provided 100% cover?
TIA Mick.

"Michael Harrington" <mikharr@bigpond.comwrites:
about 900 pick fours
I obtain 982 quadruples with this method. Please post the cover.

"Markus Triska" <triska@gmx.atwrote in message
news:87odtrspzg.fsf@gmx.at
"Michael Harrington" <mikharr@bigpond.comwrites:
>
>about 900 pick fours
>
I obtain 982 quadruples with this method. Please post the cover.

Hello Markus. Yes 980 is correct. I have a cover in 888 lines
from a Colin Barker. But he does not disclose his method.
Also I have a claim of 640 lines from a guy who discloses
neither his methods or his cover!
Let me know if you want the 888 or my generating code (VB6).

Cheers Mick.

"Michael Harrington" <mikharr@bigpond.comwrites:

I have a cover in 888 lines from a Colin Barker. But he does not
disclose his method. Also I have a claim of 640 lines from a guy
who discloses neither his methods or his cover!

640 doesn't seem unrealistic -- already the ~220 quadruples without
shared triples cover roughly half of the vertices.

Let me know if you want the 888 or my generating code (VB6).

Thanks, I was only interested in the discrepancy and thought you had
accidentally lost ~80 quadruples or used an additional heuristics.

All the best,
-- Markus Triska