search

SimonLucas / ntbea

N-Tuple Bandit Evolutionary Algorithm
13 stars 2 forks source link

Minor bug for low-dimensional spaces #6

Open SapphireBrand opened 5 years ago

SapphireBrand commented 5 years ago

The 2-tuples and 3-tuples can be impossible (which is not a problem) or overlap the n-tuples (which is a problem) if the search space size is 2 or 3.

My C++ code for this library makes n-tuples required (and eliminates the sampled points array, as recommended in issue #5 ). So my change is to only add the 2-tuples if use2Tuples && searchSpaceSize > 2. (And similar change for 3-tuples.)

SimonLucas commented 5 years ago

Thanks, I'd not thought about that! You're right, for these low dimensional spaces the duplicates should indeed be avoided.

btw: have you put your C++ version on Github?

On Sun, Feb 3, 2019 at 8:05 PM SapphireBrand notifications@github.com wrote:

The 2-tuples and 3-tuples can be impossible (which is not a problem) or overlap the n-tuples (which is a problem) if the search space size is 2 or 3.

My C++ code for this library makes n-tuples required (and eliminates the sampled points array, as recommended in issue #5 https://github.com/SimonLucas/ntbea/issues/5 ). So my change is to only add the 2-tuples if use2Tuples && searchSpaceSize > 2. (And similar change for 3-tuples.)

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/SimonLucas/ntbea/issues/6, or mute the thread https://github.com/notifications/unsubscribe-auth/ACoZ4WYtRCq-AA5gQih-6Nh1vb7NfWH0ks5vJ0EXgaJpZM4agSUk .

-- Professor Simon Lucas Head of School Electronic Engineering and Computer Science Queen Mary University of London

SapphireBrand commented 5 years ago

Not yet. I am using it for a Computer Go project, and I like how it is working. When I am convinced that it is mature enough for other people to copy/paste then I will put it up there.

In a lot of ways, the C++/template expression of the algorithm is more natural than the Java/interface expression. At least in the sense that it has fewer pieces.

From: SimonLucas notifications@github.com Sent: Monday, February 4, 2019 4:27 PM To: SimonLucas/ntbea ntbea@noreply.github.com Cc: SapphireBrand sheppardco@aol.com; Author author@noreply.github.com Subject: Re: [SimonLucas/ntbea] Minor bug for low-dimensional spaces (#6)

Thanks, I'd not thought about that! You're right, for these low dimensional spaces the duplicates should indeed be avoided.

btw: have you put your C++ version on Github?

On Sun, Feb 3, 2019 at 8:05 PM SapphireBrand <notifications@github.com mailto:notifications@github.com > wrote:

The 2-tuples and 3-tuples can be impossible (which is not a problem) or overlap the n-tuples (which is a problem) if the search space size is 2 or 3.

My C++ code for this library makes n-tuples required (and eliminates the sampled points array, as recommended in issue #5 https://github.com/SimonLucas/ntbea/issues/5 ). So my change is to only add the 2-tuples if use2Tuples && searchSpaceSize > 2. (And similar change for 3-tuples.)

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/SimonLucas/ntbea/issues/6, or mute the thread https://github.com/notifications/unsubscribe-auth/ACoZ4WYtRCq-AA5gQih-6Nh1vb7NfWH0ks5vJ0EXgaJpZM4agSUk .

-- Professor Simon Lucas Head of School Electronic Engineering and Computer Science Queen Mary University of London

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/SimonLucas/ntbea/issues/6#issuecomment-460419224 , or mute the thread https://github.com/notifications/unsubscribe-auth/AOcZXQhOdmEmmr-3vO2MZAC13m_cCMORks5vKKWSgaJpZM4agSUk . https://github.com/notifications/beacon/AOcZXeuGXLDIhyC_UqiNG0y1gB9UYkHBks5vKKWSgaJpZM4agSUk.gif

SimonLucas commented 5 years ago

Hi - I have a request for a C++ version, so would be good if you could make your C++ version public on Github. You mention you'll do it when it is mature enough - what does it lack at present?

SapphireBrand commented 5 years ago

It must be pretty good, because I use it successfully. But I am not sure that it is complete and consistent with the original.

I will push the code and we can take a look together. Give me a couple of days.

From: SimonLucas notifications@github.com Sent: Sunday, June 2, 2019 5:48 AM To: SimonLucas/ntbea ntbea@noreply.github.com Cc: SapphireBrand sheppardco@aol.com; Author author@noreply.github.com Subject: Re: [SimonLucas/ntbea] Minor bug for low-dimensional spaces (#6)

Hi - I have a request for a C++ version, so would be good if you could make your C++ version public on Github. You mention you'll do it when it is mature enough - what does it lack at present?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/SimonLucas/ntbea/issues/6?email_source=notifications&email_token=ADTRSXPVOTOGLG5UK42AAB3PYOJOVA5CNFSM4GUBEUSKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODWXR5SI#issuecomment-498015945 , or mute the thread https://github.com/notifications/unsubscribe-auth/ADTRSXOOT7ISVECO7V6QUB3PYOJOVANCNFSM4GUBEUSA . https://github.com/notifications/beacon/ADTRSXJ4Z7ZNDLSO7FMZT6LPYOJOVA5CNFSM4GUBEUSKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODWXR5SI.gif