Performance

Talk about anything Wheel Generator related which doesn't fit in the other forums
Post Reply
keymx
Newbie
Newbie
Posts: 1
Joined: Wed Feb 01, 2017 7:06 pm

Performance

Post by keymx » Wed Feb 01, 2017 7:28 pm

Even with a simple query it takes very long until a result appears.
V = 6 K = 5 T = 3 M = 3
On the activate bias it has over 1400 after about 20 min.

Base Coverage = 1, Base Coverage = 100%, Auto reduce block = Yes, Block at start = 10 is this Value importen?
Can you calculate the number of blocks? If so, how do you choose the larger number or smaller?
I thought that program calculates this value by itself and does not start at 1, which is with 15 numbers and block of 5 with 3 from 3 is not possible.

Unfortunately you have no clue as far as the program is and how long it is still running.
Is there an approximation method to calculate the runtime? I would like to check the following variants.

V = 12 to 20, K = 5 fixed, T = 3 fixed, M = 3 fixed and V = 12 to 20, K = 5 fixed, T = 2 fixed, M = 2 fixed

Thanks for your help.

User avatar
lottoarchitect
Site Admin
Posts: 1460
Joined: Tue Jan 15, 2008 5:03 pm
Location: Greece
Contact:

Re: Performance

Post by lottoarchitect » Thu Feb 02, 2017 11:44 pm

Hi keymx,
Even with a simple query it takes very long until a result appears.
I am not sure if I understand you question since observing a result is readily available right at the "best covering" menu. It may not be optimal but nevertheless, it is a result right away. So what do you mean "it takes very long"; it doesn't.
Base Coverage = 1, Base Coverage = 100%, Auto reduce block = Yes, Block at start = 10 is this Value importen?
No, it is not imported. It is just a default value, like the default values for the other parameters. The reason for not importing a value is explained next, along with
Can you calculate the number of blocks? If so, how do you choose the larger number or smaller?
I thought that program calculates this value by itself and does not start at 1, which is with 15 numbers and block of 5 with 3 from 3 is not possible.
It is virtually impossible to compute a definite and exact b parameter when we deal with NP-hard problems. Wheels optimization is such a case, NP-hard. For the simplest coverings, where no filters or other designs are needed, when we only care for the overall coverage only, it is possible to use repositories that hold current records (where you can get the b value). Such repositories are LaJolla and Wilfried's Homme but even in this case, the b indicated by those repositories may be the outcome of a specific covering theory; not the outcome of an optimization process. So for example LaJolla may indicate b=50 for a covering that has been derived via a specific optimal covering theory, however any algorithmical approach is short to achieve that same goal. Remember, we deal with NP-hard problems here. So, even if I import the value b=50 because there is such a covering available, the very next question will be, "why WG does not produce that covering at 100%, it has imported that b parameter itself so something is obviously wrong!". Nothing is wrong of course. This gets easily out of hand and gets very complicated when we introduce additional options like filters, mother coverings, dual coverage. It is impossible to estimate in advance a good value for b. We really have to define what would be considered optimal, for example, if you set the sums and odd filter in such a way that it forbids 100% coverage to be generated, there is no b value that can ever construct that 100% cover you expect. So you have to define yourself what is considered a good covering for you, and all the above are the reason the b parameter is user selectable.
Unfortunately you have no clue as far as the program is and how long it is still running.
Covering optimization is an NP-hard problem. This means, any solution found cannot be proved if it is the most optimal found, unless if it is obvious - given you define what is the optimal part- but as the optimization progresses, better solutions should be experienced. This search is open-ended. You'll never get a "job done" message because the job is never completed with NP-hard problems. The judge is the user, if the results found till that point are satisfactory, by all means stop the engine and use the outcome.

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest