[mnet-devel] analytical solution to the FEC reliability problem

[Jim McCoy]

Actually, skip the python prog (not enough precision in floats without 
a bit of work) and just feed me the numbers I need to build an excel 
spreadsheet...


On Monday, February 24, 2003, at 12:14 PM, Zooko wrote:
> Let's denote the probability that you win as "W".
>
> W = P(recover K shares) + P(recover K + 1 shares) + P(recover K + 2 
> shares) +
>    ... + P(recover all M shares)

Not sure on the formula for this one

>
> P(recover exactly K shares) = Choose(M, K) * P(S)^K * (1-P(S))^(K-M)

M=total shares
K=needed shares
P(S)=recovery of any particular share

=FACT(M)/((FACT(K)*FACT(M-K)) * P(S)^K * (1-P(S))^(K-M)

This returns a HUGE number, so obviously something is wrong here...

For 32, 16, and 0.5 we get:

choose(M, K) = 601080390
P(S)^K * (1-P(S))^(K-M) = 1.0   (hrmmm....)

>
> Now, don't forget to get the probability of recovering at *least* K 
> shares, you
> have to add up P(recover exactly I shares) for all I from K to M.

Given my results from the previous step I am not even going to try this 
one...

Jim



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
mnet-devel mailing list
mnet-devel at lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mnet-devel