Balls into Bins with Variable Sized Balls
- From: Mohammad Nabil <mohammad.nabil.h@xxxxxxxxx>
- Date: Fri, 10 Aug 2007 11:07:38 -0000
Hello,
Sorry if this is not the right group to send, I tried to seek for
statistics group but I didn't find. If this isn't the right group, I'd
hope someone would guide me to the right one. Thanks.
The classical balls into bins problem discusses the probability of
certain events occurring on throwing n balls at random on m bins. Like
the event that no bin is empty and so on. It doesn't care about the
size of the balls, however.
My problem requires that we distribute n balls among m bins ( usually
n > m, but n <= m is possible too ). Given only the mean and variance
of the balls set ( where the random variable is the size of the
ball ), I need to calculate the mean and variance of the random
variable which is the total amount of size per bin ( sum of sizes of
all the balls in some bin ), such that the variance is the minimum
possible variance.
That is I need some way to get the minimum possible variance of that
variable from some set of balls, so I'd compare it to the output of
some load-balancing algorithm.
For the input set (ball sizes set) 1 1 2 2 2 3 3 3, and n = 3, they
would be optimally distributed as 5 5 4, which has the mean 4.666667
and the (biased) variance of 0.222222. In contrast to 6 6 2 which has
the same mean and the vairance 3.555555.
At least I need some keywords to guide me where to look. I am seeking
a concrete answer, I at least want to know how to approach such
problem and solve it.
Any help is greatly appreciated.
.
- Follow-Ups:
- Re: Balls into Bins with Variable Sized Balls
- From: Mohammad Nabil
- Re: Balls into Bins with Variable Sized Balls
- Prev by Date: Re: Definition Of Sanity - {HRI 20040410-V2.0.1}
- Next by Date: Re: Balls into Bins with Variable Sized Balls
- Previous by thread: Re: Definition Of Sanity - {HRI 20040410-V2.0.1}
- Next by thread: Re: Balls into Bins with Variable Sized Balls
- Index(es):
Relevant Pages
|