in reply to Re: Getting Binomial Distribution under Math::Pari (log) and Combinatorial Method
in thread Getting Binomial Distribution under Math::Pari (log) and Combinatorial Method

When speed matters (apart from ability to deal with *BIG* number), the performance are:
Rate binomial_log binomial_pari binomial_comb binomial_log 6164/s -- -61% -87% binomial_pari 15709/s 155% -- -66% binomial_comb 45985/s 646% 193% --
Regards,
Edward
  • Comment on Re^2: Getting Binomial Distribution under Math::Pari (log) and Combinatorial Method
  • Download Code

Replies are listed 'Best First'.
Re^3: Getting Binomial Distribution under Math::Pari (log) and Combinatorial Method
by tilly (Archbishop) on Mar 24, 2005 at 03:08 UTC
    If you just want an approximation, then performance may be better still if you use the fact that from Stirling's formula the log of n! is approximately
    log(n**n * exp(-n) * sqrt(2*PI*n)) = n*log(n) - n + (log(2 + log(PI) + log(n))/2
    Plug that into the fact that n choose m is n!/(m!*(n-m)!) and you can come up with a good approximation that uses Perl's native arithmetic.

    This approximation may well turn out to be the fastest approach for large numbers.

    Update: From Mathworld I just found out that a significantly better approximation of n! is sqrt(2*n*PI + 1/3)*n**n/exp(n). Algebraically this is less convenient, but the improved accuracy may matter for whatever you're trying to do.