Re: Need your help again to clarify output, how to get reduced variables & value ?

Hello,

> Since you have moved on from the previous 500 constraint problem to
this new
> 900 constraint problem, does that mean you were able to solve it?
How?
> What is the class of problems these come from?. I had guessed
something like
>; 2-designs.

Many thanks for hints where to look into source code !
I must take some time to look there.

What is a 2-design ? "1-design" and "2-design" is a bad word for
google, a lot of wrong hits...

In general:

I learned that i can use "bin" instead of "int". And meanwhile i can
use the API for my first steps instead of the IDE. But i still don't
know what all the options mean in IDE (e.g. different algorithms and
parameters). Is there perhaps some introduction ?

With these above changes i am able to at least load and reduce my
sample lp files without any crashes. But i think i am far away from
solving them. At least i think it is trying to solve my lp (cpu usage
goes up), but i have no sign if there is any progress at all or if i
need 1000th of years till end or it's done the next second...

My problem is some kind of sudoko, but much bigger. For each
possibility i have a boolean variable. So a lot of my variable are 0,
but a few (e.g. 256 out of 130000) of them should be 1 in the
solution. I don't know yet, if the use of lp_solve can help me at
all, but i find it interesting how lp_solve can do this. I had some
basic introduction into linear optimization (calculating easy
problems by hand) at university (long time ago), but i newer used a
computer to solve such problems before.

What algorithm and parameter are important or should i choose / look
for when using big boolean problems ? And is there some kind of
estimation, e.g. 1000 variable problem can be solved in an hour. But
10000 variable problems can't be solved in life time ? Or is it more
the ratio variable : constraints ?

Many thanks for your help !

Regards,

Martin

Messages in this topic (3) Reply (via web post) | Start a new topic

.

"Is there perhaps some introduction?"
First the help file *.chm or online http://lpsolve.sourceforge.net/5.5/
Try the set_bb_rule and other routines with names from the BB section of the
IDE.
I think the comments near the definitions of the Branch and bound option
defines in
Lp_lib.h line 493 /* B&B strategies */ ; not so good so the code elsewhere.
Say look for NODE_GREEDYMODE leads to reading lp_mipbb.c
/* Node and branch management routines */
STATIC MYBOOL initbranches_BB(BBrec *BB)

================================ 2-designs
Designs and association schemes.

http://www.iasri.res.in/ebook/EBADAT/2-Basic%20Statistical%20Techniques/10-I
BD-vksharma.pdf
http://www.iasri.res.in/iasriwebsite/DESIGNOFEXPAPPLICATION/Electronic-Book/
Module%202/3PBIBD.pdf
The algebraic approach involving character theory is shown in this course.
http://www.maths.qmul.ac.uk/~rab/MAS417/
http://www.maths.qmul.ac.uk/~pjc/preprints/gendesdm.pdf

Lots of other names here orthogonal Latin squares, Steiner triple systems
...

Why not start small. Look at SUDUKO solution by LP
Google lp solver for sudoku

Here is a link with an lp_solve model generator
http://www.sudoku.com/boards/viewtopic.php?p=720&sid=e184d2f58db156c35e55599
530d42fe6

William
________________________________________
From: lp_solve%40yahoogroups.com">lp_solveyahoogroups.com [mailto: lp_solve%40yahoogroups.com">lp_solveyahoogroups.com] On Behalf
Of capiman26061973
Sent: Sunday, June 15, 2008 11:32 AM
To: lp_solve%40yahoogroups.com">lp_solveyahoogroups.com
Subject: [lp_solve] Re: Need your help again to clarify output, how to get
reduced variables & value ?

Hello,

> Since you have moved on from the previous 500 constraint problem to
this new
> 900 constraint problem, does that mean you were able to solve it?
How?
> What is the class of problems these come from?. I had guessed
something like
>; 2-designs.

Many thanks for hints where to look into source code !
I must take some time to look there.

What is a 2-design ? "1-design" and "2-design" is a bad word for
google, a lot of wrong hits...

In general:

I learned that i can use "bin" instead of "int". And meanwhile i can
use the API for my first steps instead of the IDE. But i still don't
know what all the options mean in IDE (e.g. different algorithms and
parameters). Is there perhaps some introduction ?

With these above changes i am able to at least load and reduce my
sample lp files without any crashes. But i think i am far away from
solving them. At least i think it is trying to solve my lp (cpu usage
goes up), but i have no sign if there is any progress at all or if i
need 1000th of years till end or it's done the next second...

My problem is some kind of sudoko, but much bigger. For each
possibility i have a boolean variable. So a lot of my variable are 0,
but a few (e.g. 256 out of 130000) of them should be 1 in the
solution. I don't know yet, if the use of lp_solve can help me at
all, but i find it interesting how lp_solve can do this. I had some
basic introduction into linear optimization (calculating easy
problems by hand) at university (long time ago), but i newer used a
computer to solve such problems before.

What algorithm and parameter are important or should i choose / look
for when using big boolean problems ? And is there some kind of
estimation, e.g. 1000 variable problem can be solved in an hour. But
10000 variable problems can't be solved in life time ? Or is it more
the ratio variable : constraints ?

Many thanks for your help !

Regards,

Martin

Messages in this topic (4) Reply (via web post) | Start a new topic

.