Re: Low hanging fruit - Accelerated random distribution functions

[Paul Kienzle]

On Feb 23, 2007, at 7:49 AM, Paul Kienzle wrote:

> On Feb 23, 2007, at 5:27 AM, David Bateman wrote:
>
> > Paul Kienzle wrote:
> > > On Feb 22, 2007, at 7:29 PM, Paul Kienzle wrote:
> > >
> > > > On Feb 22, 2007, at 11:46 AM, David Bateman wrote:
> > > >
> > > > > I haven't tried to convert the functions geornd,
> > > >
> > > > Octave uses the following; it would be hard to beat:
> > > >
> > > >    rnd(k) = floor (log (rand (size (k))) ./ log (1 - p(k)));
> > >
> > > On second thought,
> > >
> > >     rnd = floor(-rande(n)) ./ log(1-p)

Oops ... misplaced parenthesis...

It should be R = floor(-rande(n)./log(1-p))

I'm not sure why this would affect the speed.

> > > is 2x faster.
> >
> > Are you sure its not the reverse
> >
> > octave:3> function rnd = geornd2(p,varargin), rnd =
> > floor(-rande(varargin{:})) ./ log(1-p); endfunction
> > octave:4> tic; b0 = geornd(3,1,1e7); toc
> > Elapsed time is 1.643173 seconds.
> > octave:5> tic; b1 = geornd2(3,1,1e7); toc
> > Elapsed time is 3.176453 seconds.
>
> octave2.9:10> tic; y = floor(log(rand(400))./log(1-0.2)); toc
> Elapsed time is 0.534271 seconds.
> octave2.9:11> tic; x = floor(-rande(400)./log(1-0.2)); toc
> Elapsed time is 0.348267 seconds.
> octave2.9:12> [statistics(x(:)),statistics(y(:))]
> ans =
>
>     0.00000    0.00000
>     1.00000    1.00000
>     3.00000    3.00000
>     6.00000    6.00000
>    55.00000   54.00000
>     4.00524    3.99577
>     4.49588    4.46590
>     2.02001    1.99510
>     5.99429    5.90314