2009/9/16 Jason Grout <jason-s...@creativetrax.com>:
>
> William Stein wrote:
>
>> I did write a lot of stats in Cython already, and it's much faster
>> than both R and scipy.stats at what it does (at least last time I
>> checked). This is all the code in Sage's finance.TimeSeries...
>> It's very specialized though compared to what is offered by Scipy/R.
>> Anyway, the one time I seriously wanted to do a lot of stats, I
>> found scipy.stats and R to both be way way too slow, so had to write
>> my own in Cython.
>
>
> Do you know how your code compared to gsl in speed?
No.
> Were you talking to R via pexpect?
No but it wouldn't matter. The main thing I benchmarked was
computing the standard deviation of about a million (or more) numbers.
I spent hours tracking down why R was much slower (at least 10
times) than what anybody could implement in 5 minutes in Cython, and
it was because so far as I can tell (at the time) R's standard
deviation was implemented in terms of a general n-dimensional
covariance (?) function. The function was so general that it slowed
down a lot compared to naive code in the special case of 1-dimensional
data. I looked at numpy and it was also slower for much the same
reason. This could have all changed by now, as this was about 2
years ago.
I was also very interested in generating many numbers at random that
are normally distributed.
William
>
> Jason
>
> --
> Jason Grout
>
>
> >
>
--
William Stein
Associate Professor of Mathematics
University of Washington
http://wstein.org
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to
sage-devel-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---
