Re: Multiplicative Chernoff bound
- From: ALiX <alix.tofigh@xxxxxxxxx>
- Date: 18 Apr 2007 01:56:58 -0700
I got this as a special case from the inequality given in
http://en.wikipedia.org/wiki/Chernoff_bounds.
Using the chernoff bound you mentioned I came up with a procedure. I
would be grateful for any thoughts about it.
As before I have a random process whose outcome is either success or
failure. I am sampling (independant trials) to estimate the
probability p of success. First I decide on some confidence level, say
c = 0.9. While sampling I keep track of how many time I have sampled,
and the average number of failures and successes m per sample. I use
the chernoff bound to find e such that with probability c, m-e < p < m
+e. I stop when this interval is small enough to ensure that the
relative error of my estimation m of p is as small as I want (with
probability c). What do you think?
/ALiX
.
- Follow-Ups:
- Re: Multiplicative Chernoff bound
- From: Herman Rubin
- Re: Multiplicative Chernoff bound
- From: Bo
- Re: Multiplicative Chernoff bound
- References:
- Multiplicative Chernoff bound
- From: ALiX
- Re: Multiplicative Chernoff bound
- From: Jack Tomsky
- Multiplicative Chernoff bound
- Prev by Date: Re: Selling hundreds of solutions manuals
- Next by Date: Re: converting Exceedance probability into a probability distribution.
- Previous by thread: Re: Multiplicative Chernoff bound
- Next by thread: Re: Multiplicative Chernoff bound
- Index(es):
Relevant Pages
|
