[Clp] CLP and COIN parameters/ was corrections to coinmp
Rod Frowd
rod at frowd.net
Wed Feb 20 10:24:33 EST 2008
Larry:
You are correct about the scaling parameter.
When I started this I was using the pre-built dll and after some more
testing it seems that most of these parameter changes are unnecessary
when using the dll built from the latest source. As far as I can tell
the prebuilt dll was built 2-3 years ago and it seems that there have
been many changes to the baseline code since then.
The only parameter change that seem to make a difference to the problem
I am solving with the dll built from latest source is the perturbation
parameter.
Rod
Larry A. Taylor wrote:
> At 05:51 PM 2/18/2008, Rod Frowd wrote:
>> Parameters I changed were:
>>
>> COIN_INT_SCALING set to 5 was 3
>> COIN_INT_PERTURBATION set to 50 was 100
>> COIN_INT_PRIMALPIVOTALG set to 4 was 1
>> COIN_INT_DUALPIVOTALG set to 3 was 1
>>
>> It went from 15000 iterations to around 5000 with these changed. The
>> problem I am solving is not well conditioned numerically so these
>> parameters seem to help.
>> ...
>> Rod
>
> I have been looking at various places in the code, and I thought that
> the only
> options to COIN_INT_SCALING were 0,1,2,3 and 4.
>
> For instance, in CoinMP.cpp, the options table has these lines:
>> "Scaling" , "Scaling" ,
>> GRP_SIMPLEX, 3, 3, 0, 4, OPT_LIST, 0,
>> COIN_INT_SCALING,
>> /* Sets or unsets scaling, 0 -off, 1 equilibrium, 2 geometric,
>> 3, auto, 4 dynamic(later) */
>
> ... and there is a similar comment in Clp_C_Interface.cpp.
>
> So where is option 5 documented, and what does it do?
>
> LAT
-------------- next part --------------
A non-text attachment was scrubbed...
Name: rod.vcf
Type: text/x-vcard
Size: 302 bytes
Desc: not available
URL: <http://list.coin-or.org/pipermail/clp/attachments/20080221/6ffb3e80/attachment.vcf>
More information about the Clp
mailing list