[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: a list of manually fine-tuned beaming exceptions?
From: |
Janek Warchoł |
Subject: |
Re: a list of manually fine-tuned beaming exceptions? |
Date: |
Sun, 6 Mar 2011 00:18:17 +0100 |
2011/3/6 address@hidden <address@hidden>:
>
> On Mar 5, 2011, at 17:33, Janek Warchoł <address@hidden> wrote:
>> Thanks for information! Yes, this should be investigated.
>> However, i'm afraid that when i change some well-established
>> parameters, it may cause unwanted side-effects in other beams. And the
>> worst part is that i have no means to check this: it would require a
>> gigantic proof-sheet consisting of thousands of different beams to
>> check that some parameter combination works optimal.
>> That's why i suggested a temporary solution. It could also serve as a
>> reference beaming database later.
>
> Having recently OD'd on beam quanting, I am convinced that the right cocktails
> of penalties and demerits can hit all of the targets that would potentially
> be on this list.
> There's nothing wrong with keeping a list like this in the form of a regtest
> with lots of beams (I think it's a great idea!), but I think the goal should
> be to modify
> the quanting such that it attains the result you want while not breaking
> other regtests.
> I know this is a tall order and means more work, but it seems like a better
> use
> of contributor time and a more sustainable solution.
Ok. It's definately beyond me now, so i'll come back to it later.
Meanwhile i'll collect samples for the regtest - please send me
snippets if you find beams looking bad.
I'll speak again when i reach 200 examples :)
cheers,
Janek