Random number generation from functions

[drs]

Is there any way to generate random numbers based on arbitrary real valued
functions? I am looking for something like random.gauss() but with natural
log and exponential functions.

thanks,

-d

[Bengt Richter]

Re: Random number generation from functions

On Mon, 29 Nov 2004 20:51:50 GMT, "drs" <drs@remove-to-send-mail-ecpsoftware.com > wrote:
[color=blue]
>Is there any way to generate random numbers based on arbitrary real valued
>functions? I am looking for something like random.gauss() but with natural
>log and exponential functions.
>
>thanks,[/color]
Don't know what you mean. This kind of thing?
(the 1-random... is to make the random numbers in (0,1] instead of [0,1))
[color=blue][color=green][color=darkred]
>>> import random
>>> import math
>>> math.log(1.0-random.uniform( 0,1))[/color][/color][/color]
-0.9003028845584 1156[color=blue][color=green][color=darkred]
>>> math.log(1.0-random.uniform( 0,1))[/color][/color][/color]
-0.2612491418648 35[color=blue][color=green][color=darkred]
>>> math.log(1.0-random.uniform( 0,1))[/color][/color][/color]
-0.9969436681854 7997

Regards,
Bengt Richter

[Colin J. Williams]

Re: Random number generation from functions

drs wrote:[color=blue]
> Is there any way to generate random numbers based on arbitrary real valued
> functions? I am looking for something like random.gauss() but with natural
> log and exponential functions.
>
> thanks,
>
> -d
>
>[/color]
numarray has a random package which provides a number of functions,
including: normal( mean, stddev, shape=[])

That may serve your need.

Colin W.

[CptPicard]

Re: Random number generation from functions

drs wrote:[color=blue]
> Is there any way to generate random numbers based on arbitrary real valued
> functions? I am looking for something like random.gauss() but with natural
> log and exponential functions.
>
> thanks,
>
> -d
>
>[/color]
I remember for having used it on a gaussian generator that you can do that very
easily by yourself:
- you generate a random real number in the range [0 .. 1],
- then you apply an inverse probability density function (e.g. inverse gaussian
pdf : http://www.itl.nist.gov/div898/softw...ar/rigpdf.htm),
and you have then your random number following the specified law (here gaussian).

[Robert Kern]

Re: Random number generation from functions

drs wrote:[color=blue]
> Is there any way to generate random numbers based on arbitrary real valued
> functions? I am looking for something like random.gauss() but with natural
> log and exponential functions.[/color]

scipy[1] has a large collection of "standard" univariate pdfs, including
normal, exponential, gamma, and the like.

One cannot generate random numbers from arbitrary real valued functions
because a function needs to satisfy certain restrictions to be a pdf.
For example, you cannot generate random numbers "with a log function",
unless you restrict the domain and renormalize.

Generating random numbers from an arbitrary pdf is possible, but tricky.
I encourage you to search the literature for "monte carlo sampling."

[1] http://www.scipy.org

--
Robert Kern
rkern@ucsd.edu

"In the fields of hell where the grass grows high
Are the graves of dreams allowed to die."
-- Richard Harter

[Alejandro López-Valencia]

Re: Random number generation from functions

On Mon, 29 Nov 2004 20:51:50 GMT, "drs"
<drs@remove-to-send-mail-ecpsoftware.com > wrote:
[color=blue]
>Is there any way to generate random numbers based on arbitrary real valued
>functions? I am looking for something like random.gauss() but with natural
>log and exponential functions.[/color]

Try with CRNG, it may have what you need, or be able to adapt it to
what you want. http://www.sbc.su.se/~per/crng/