[Numpy-discussion] in place random generation

[Anne Archibald]

On 08/03/07, Charles R Harris <charlesr.harris at gmail.com> wrote:

> For normals this seems overkill as the same result can be achieved by an
> offset and scale, i.e., if r is an array of random numbers with mean 0 and
> sigma 1, then
>
> myrandomarray = (r*mysigma + mymean)
>
> easily achieves the same result. Other distributions don't have such happy
> properties, unfortunately, and will have high overhead regardless. For
> instance, Poisson distributions require a computation of new internal
> parameters for each value of the mean and doing this on an item by item
> basis over a whole array is a terrible idea. Hmm, I am not convinced that
> broadcasting is going to buy you much except overhead. Perhaps this problem
> should be approached on a case by case basis rather than by some global
> scheme.

Whether it's efficient or not in terms of CPU time, it's extremely
handy to be able to do
photons = numpy.random.poisson(0.1*cos(omega*numpy.arange(1000000)+0.9)

Of course I'd prefer if it ran faster, but many of numpy's operations
are notably slower than the same operation on (say) python lists (try
the Programming Language Shootout sometime...); their real advantage
is that they allow programs to be written quickly and clearly.

Anne