Login / Register  0 items | $0.00 NewWhat is KVR? Submit News Advertise
User avatar
Aleksey Vaneev
KVRAF
 
3212 posts since 7 Sep, 2002

Postby Aleksey Vaneev; Sat Jun 17, 2017 11:56 am Re: Bitmask evolution optimization method from Aleksey Vaneev

CurryPaste wrote:Thank you for posting your research here, Aleksey.

I know its not intended for real-time use, but do you think its approaching something nearly useful for simple real-time optimisation problems? Im thinking perhaps some optimisation problems which are local enough to keep the amount of iterations low.

I do not see a problem with that as method's overhead is relatively low. Generating seveal solutions per second should not be a problem. For real-time audio? I'm not so sure.
Image
Squidsneeze
KVRist
 
39 posts since 3 Nov, 2015, from Germany

Postby Squidsneeze; Sat Jun 17, 2017 12:31 pm Re: Bitmask evolution optimization method from Aleksey Vaneev

Hey Aleksey,

just a small remark to your code: data allocated with new[] should be released with delete[]. But your "deleteBuffers" function just calls delete which results in undefined behavior and a possible memory leak ^^ If that was intentional for some reasons, a small comment in the code would clear up things.

Best,
Danny
Miles1981
KVRian
 
1208 posts since 26 Apr, 2004, from UK

Postby Miles1981; Sat Jun 17, 2017 2:53 pm Re: Bitmask evolution optimization method from Aleksey Vaneev

Even better would be to have std::unique_ptr<[]> for the parameters instead of raw pointers...
User avatar
Aleksey Vaneev
KVRAF
 
3212 posts since 7 Sep, 2002

Postby Aleksey Vaneev; Sun Jun 18, 2017 12:02 am Re: Bitmask evolution optimization method from Aleksey Vaneev

Squidsneeze wrote:Hey Aleksey,

just a small remark to your code: data allocated with new[] should be released with delete[]. But your "deleteBuffers" function just calls delete which results in undefined behavior and a possible memory leak ^^ If that was intentional for some reasons, a small comment in the code would clear up things.

Best,
Danny

Thanks, that was a mistake.
Image
User avatar
Aleksey Vaneev
KVRAF
 
3212 posts since 7 Sep, 2002

Postby Aleksey Vaneev; Tue Jun 20, 2017 6:37 am Re: Bitmask evolution optimization method from Aleksey Vaneev

Ha ha, I've evolved the method beyond "bitmask evolution". It now uses a different approach to parameter randomization. The name is now outdated, but for historic reasons the name will remain as "biteopt". A year-long path of method's evolution.
https://github.com/avaneev/biteopt

Many improvements implemented. So far the method can't solve only about 8% of test functions I found so far. Still trying to figure out how to improve the method.
Image
Previous

Moderator: Moderators (Main)

Return to DSP and Plug-in Development