[SciPy-Dev] seeded randn gets different values on osx

Pauli Virtanen pav@iki...
Fri Jun 18 03:09:25 CDT 2010


Sun, 13 Jun 2010 16:16:56 -0600, Charles R Harris wrote:
[clip]
> I've been thinking that the ziggurat implementation should be added as a
> separate function, znormal or zignomal, perhaps together with a fast
> version returning lower precision that could be helpful in simulating
> noise.

If the only difference in practice is that the ziggurat returns a 
different sequence of numbers for a given seed, then perhaps just a 
ziggurat=True keyword argument could be used for the algorithm selection?

-- 
Pauli Virtanen



More information about the SciPy-Dev mailing list