Thread: CRT offsets
View Single Post
Old 2021-04-05, 13:51   #11
R. Gerbicz
R. Gerbicz's Avatar
"Robert Gerbicz"
Oct 2005

27268 Posts

Originally Posted by henryzz View Post
That's a nice improvement. How did you find this?
For that record used only your 1st method with the following modification:
collect in array/vector those res values that occur maximal times as x%p and choose randomly(!) one res value from these. It gives some breath for the algorithm, but notice that you can stuck in a local min, so without several restarts you could be far from global min. I'd say the above result is not very far from optimal, but this is just my guess.

Originally Posted by henryzz View Post
I think this shows that it is necessary to pay attention to the target gap size with these searches.
Yeah, use it if you'd want to reach say merit=25, use a larger gap for merit=30. Ofcourse we don't know in advance what merits you will get in a search, but using this gap for merit=40 isn't that useful/optimal.
R. Gerbicz is offline   Reply With Quote