Help with covering these 2 parameters

Talk about anything Wheel Generator related which doesn't fit in the other forums
Post Reply
magnifico
Casual
Casual
Posts: 13
Joined: Fri Aug 16, 2019 4:31 pm

Help with covering these 2 parameters

Post by magnifico » Sat Aug 17, 2019 10:07 am

Hi, I hope you can help me with finding the best settings for covering these 2 parameters: 18 6 3 6 with 7 tickets ( World Record) and 20 6 3 6 with 10 tickets ( World Record). The reason I'm asking for your help is because no meter what optimization settings I try, WG doesn't seem to cover these parameters. Once I did mange to cover the 18 6 3 6 with 7 tickets using WG version 1.8.5 with these settings:
- search strength: Extensive + Cyclic and Full Cyclic
- improvement criteria: Proportional
- Bias Expansion: To the maximum right
- Bias Tension: To the maximum right
- Additional memory available: 512 MB
- Enable special covering speedup: ON
That's it. But I only manage to achieve this once. If I tighten the settings, things get even worse. It will go to about 99.31% cover than it just stops there, no matter how much I let it run (by the way I've let it run, for each settings, about 24 hours). If I use the old DOS program lotto ININUGA, these 2 parameters are covered in about 15-20 minutes. So, am I doing something wrong ? Am I missing some hidden settings ?

Thank you, in advance, for your help.

PS: I apologize for possible grammar mistakes. English is not my native language.

User avatar
lottoarchitect
Site Admin
Posts: 1635
Joined: Tue Jan 15, 2008 5:03 pm
Location: Greece
Contact:

Re: Help with covering these 2 parameters

Post by lottoarchitect » Sun Aug 18, 2019 11:38 am

The problem with non-deterministic approaches is that a particular set of settings will not go through the exact same steps as it did at a particular run. That means you may get a result in X time, if you run the exact same settings you may get that result in more or less time, or not at all, or get an ever better result. Most of the time, if the settings are sort of appropriate for the required result you'll get the desired outcome sooner or later.

WG is probably capable to make these coverings given they are not pack designs and their particular size. I would surely avoid extreme settings for expansion and tension, not sure why you have moved them to the far right?? Improvement criteria:absolute, no use of proportional here as you don't involve extra filters. Apart from that, my approach would generally be:

Bias expansion/tension: 0/0 or 0/-1000 or 1000/-1000 depending on how often it approaches the best coverage found (we want it to occasionally reach it)
Strength: standard
2048 MB + special speedup ON
Impr. criteria : absolute

and just let it work on the coverings. I can't tell if it will make it faster of within 1 hour but I do expect it to make these at some point. Ininuga probably uses some other approach for building, it may happen to produce quicker results in these particular coverings, probably WG can produce quicker results in other coverings too. The important in WG's optimization is to see the current coverage occasionally reach the current best found. Finally, not all coverings can be made to their absolute current records, especially if they are derived from covering theory (like pack/steiner designs) but I don't expect these particular coverings to be that case really. I'd gradually reduce the bias if I don't see something improving after a few hours, i.e. from 0 to -200, then -400 etc.

To the more technical info, these coverings probably have very few solutions to their current record, therefore it takes quite much time to find these solutions. It doesn't matter the size of the covering or if it is considered small or big, if the available solutions to be found is small, it will certainly take quite more time to find one compared to a bigger covering that its current record can be picked among many more solutions. The difficulty to find easily these solutions in those coverings even if they look like small coverings (7 or 10 blocks), in reality suggests the available solutions are quite few.

magnifico
Casual
Casual
Posts: 13
Joined: Fri Aug 16, 2019 4:31 pm

Re: Help with covering these 2 parameters

Post by magnifico » Sun Aug 18, 2019 12:35 pm

Thank you for your answer. I forgot to mention :oops: Indeed Ininuga program uses besides it's normal "run"mode, a special mode called "cascade". In this special mode it reaches the mentioned constructions. Maybe you could take a look at it. It's free and who knows you might find something useful for improving WG wheeling system. ;) As for your tips I've always wondered something about "improvement criteria" section. I've red the help file and I understand that absolute mode should be used with filters off, and proportional with filters on. But if I use proportional with filters off, isn't just like using the absolute mode? In the help file, couldn't find the difference between "absolute" mode and "proportional/filters off" mode. Maybe you could offer us some details. And another question: In the progress/improvements section a (b/c), can a higher "b" number be an indicator that we are on the right track of covering the parameters, as long as we respect the other criteria you've suggested in the help file ("b" should be at most 10% of "c", "c" should not be above 100000 and "a" saw no improvements etc.) From what I've read the indicator "b" represents the number of times the current thread has hit the best current cover (hope I got it right :roll: ). Thank you.

PS: Only with those extreme bias settings wg 1.85 covered the parameters in 7 tickets and under 5 hours. It probably found a lucky working thread that lead to a solution.

Post Reply

Who is online

Users browsing this forum: No registered users and 7 guests