On the "Reken mee met ABC" sister site I noticed that there is an overview of the quality of triples found, and even a ranking per user.
I have following questions:
* How do you define the quality of a triple?
* How come a single user can find the best quality triple if a work unit is handled by at least 2 users?
____________
BOINC.BE: For Belgians who love the smell of glowing red cpu's in the morning
Tutta55's Lair
/ 
* How come a single user can find the best quality triple if a work unit is handled by at least 2 users?
i would guess that the canonical result gets the fame lol
____________
/ The quality of (a,b,c) is log(c)/log(rad(abc)). On the site it should be explained
somewhere what that means.
The first result forms the canonical wu id and 'finds' it. Over all being first and second equal out.
/ The quality of (a,b,c) is log(c)/log(rad(abc)). On the site it should be explained somewhere what that means.
could you help out the non dutch speaking(reading) people please
____________
/ 
The first result forms the canonical wu id and 'finds' it. Over all being first and second equal out.
You mean that the first result of two gets the price?
/
The first result forms the canonical wu id and 'finds' it. Over all being first and second equal out.
You mean that the first result of two gets the price?
thats the way i am reading it reduced cache from 3 days to one hour lol
____________
/
The first result forms the canonical wu id and 'finds' it. Over all being first and second equal out.
You mean that the first result of two gets the price?
thats the way i am reading it reduced cache from 3 days to one hour lol
So its not better to go to one result only?
/
The first result forms the canonical wu id and 'finds' it. Over all being first and second equal out.
You mean that the first result of two gets the price?
thats the way i am reading it reduced cache from 3 days to one hour lol
So its not better to go to one result only?
i guess project needs to confirm that results are correct and i can agree with that since this project will run for quite some time and results will effect all other science since math is the underlying language of other science
____________
/
i guess project needs to confirm that results are correct and i can agree with that since this project will run for quite some time and results will effect all other science since math is the underlying language of other science
Thats true but so the fastest CPU wins ever ^^
/
i guess project needs to confirm that results are correct and i can agree with that since this project will run for quite some time and results will effect all other science since math is the underlying language of other science
Thats true but so the fastest CPU wins ever ^^
i am thinking that the cpu with the smaller cache wins for example i was using a 3 day cache going though and finding what work units had been returned and only crunching those work units in an effort to stay in the top 20, which was never going to happen since i only have to comps (athlon 1500+ amd64x2 4600+) which has been fun but i am favouring fame vs credits (or if you like being in the top 20) in other word i am waving the white flag so to speak or get another 10 or so amdx2 4600+ lol like my wife would let me do that lol
ps does any one know much about beowulf clusters ?
edit changed , to l
____________
/
The first result forms the canonical wu id and 'finds' it. Over all being first and second equal out.
You mean that the first result of two gets the price?
thats the way i am reading it reduced cache from 3 days to one hour lol
So its not better to go to one result only?
It doesn't matter if you complete it first.
I know it's a bit fishy with a quorum greater than 1, but we want of course
reliability of the data. And having a n-multiple of names everywhere
if there's a quorum of n is also something I don't want.
Yuckup, about the quality of a number, it is q = log(c)/log(rad(abc)) where rad
is the radical, i.e. the prime decomposition of abc with all exponents set
equal to 1. If q > 1 then a+b=c is an abc-triple, or a hit. Those are the ones
we want to find. The abc-conjecture implies that there are only finitely many
such a,b,c with quality greater than any number > 1. If this is true?
Who knows :)
/ Over all being first and second equal out.
If nobody will ever know you raced, it becomes a different game (see below).
You mean that the first result of two gets the price?
And having a n-multiple of names everywhere
if there's a quorum of n is also something I don't want.
I highly recommend you change that policy Hendrik, MANY people will be disappointed by this, yet, undisclosed 'detail'.
zero-g.
/ Over all being first and second equal out.
If nobody will ever know you raced, it becomes a different game (see below).
You mean that the first result of two gets the price?
And having a n-multiple of names everywhere
if there's a quorum of n is also something I don't want.
I highly recommend you change that policy Hendrik, MANY people will be disappointed by this, yet, undisclosed 'detail'.
zero-g.
It is not a policy as it is more or less designed by boinc, having
a canonical wu. It does not become a 'different game', statistically
everybody has the same chances to find results given that they
compute evenly. I could with effort track down everybody
that computed something, but it will make things more complicated,
and everything slower and twice as much work has to be done when generating
stats at the least.
We're going to face already difficulties soon with tracking everything and recording everything with space, because it is a lot of data.
That is why I didn't chose to do it that way, I really don't want
to make it even more complicated.
/