Re: [algogeeks] Re: BiasedRandom fron unbiasedRandom.

2010-10-16 Thread nishaanth
@snehal...a simple way to do it is..

create an array of lets say 1000...

fill in 1000* p elements with 0 and the rem with 1

now use rand() and generate the index..and return arr[index]

On Fri, Oct 8, 2010 at 8:49 PM, Dave  wrote:

> @Snehal: If p has a finite binary representation, of say n bits, then
> generate a sequence of up to n unbiased random numbers, continuing the
> sequence as long as the ith random number agrees with the ith bit to
> the right of the binary point. When the ith number disagrees with the
> ith bit, return that ith number. If the computation continues until
> even the nth number matches the nth bit, then start over.
>
> Example: Suppose that the binary representation of p is 0.101101.
> If the first unbiased random number is 1, generate the second number.
> If the second number is 0, continue to the third number.
> If the third number is 0, return 0.
>
> If p does not have a finite binary representation, e.g., if p = 1/3 or
> sqrt(1/2) and you are not using a finite precision floating point
> representation, then the algorithm will terminate with probability 1.
>
> Dave
>
> On Oct 8, 2:44 am, snehal jain  wrote:
> > Given a unbiased function that generates 0 and 1 with equal
> > probability write a function biasedrandom that genreates 0 with
> > probability p and 1 with probability 1-p.
>
> --
> You received this message because you are subscribed to the Google Groups
> "Algorithm Geeks" group.
> To post to this group, send email to algoge...@googlegroups.com.
> To unsubscribe from this group, send email to
> algogeeks+unsubscr...@googlegroups.com
> .
> For more options, visit this group at
> http://groups.google.com/group/algogeeks?hl=en.
>
>


-- 
S.Nishaanth,
Computer Science and engineering,
IIT Madras.

-- 
You received this message because you are subscribed to the Google Groups 
"Algorithm Geeks" group.
To post to this group, send email to algoge...@googlegroups.com.
To unsubscribe from this group, send email to 
algogeeks+unsubscr...@googlegroups.com.
For more options, visit this group at 
http://groups.google.com/group/algogeeks?hl=en.



[algogeeks] Re: BiasedRandom fron unbiasedRandom.

2010-10-08 Thread Dave
@Snehal: If p has a finite binary representation, of say n bits, then
generate a sequence of up to n unbiased random numbers, continuing the
sequence as long as the ith random number agrees with the ith bit to
the right of the binary point. When the ith number disagrees with the
ith bit, return that ith number. If the computation continues until
even the nth number matches the nth bit, then start over.

Example: Suppose that the binary representation of p is 0.101101.
If the first unbiased random number is 1, generate the second number.
If the second number is 0, continue to the third number.
If the third number is 0, return 0.

If p does not have a finite binary representation, e.g., if p = 1/3 or
sqrt(1/2) and you are not using a finite precision floating point
representation, then the algorithm will terminate with probability 1.

Dave

On Oct 8, 2:44 am, snehal jain  wrote:
> Given a unbiased function that generates 0 and 1 with equal
> probability write a function biasedrandom that genreates 0 with
> probability p and 1 with probability 1-p.

-- 
You received this message because you are subscribed to the Google Groups 
"Algorithm Geeks" group.
To post to this group, send email to algoge...@googlegroups.com.
To unsubscribe from this group, send email to 
algogeeks+unsubscr...@googlegroups.com.
For more options, visit this group at 
http://groups.google.com/group/algogeeks?hl=en.