commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Phil Steitz" <>
Subject Re: [math] Random data and lang.math.RandomUtils
Date Tue, 13 May 2003 14:52:36 GMT
robert burrell donkin wrote:
> a few observations:
> 1. there isn't really any harm in (some) duplication between math and lang.
> 2. what lang wants to include is an issue best left to the lang component.
> 3. there isn't any reason why stuff can't be refactored later. it's 
> easier to discuss code in CVS.

Thanks.  I will submit this with minimal refactoring and we can talk 
about it in concrete terms.


> - robert
> On Tuesday, May 13, 2003, at 02:49 AM, Phil Steitz wrote:
>> I am pulling together another submission that includes the random data 
>> generation stuff that sent us down this path and I need to decide what 
>> to submit as patches to lang's RandomUtil's implementation and what to 
>> include here.  Here is my current plan:
>> nextIntBetween(lower,upper), nextLongBetween(lower,upper), 
>> nextGaussian(mu,sigma) go to lang.math.RandomUtils.
>> nextExponential(lambda), nextPoisson(lambda) and some other parametric 
>> generators are included in a new math.DataGeneration class that 
>> exposes all of these (using the lang implemention for uniform, 
>> Gaussian deviates)
>>  as well as the following:
>> * random values generated using an empirical probability distribution
>>   estimated from an input file (i.e., values distributed "like" those in
>>   the input file)
>> * random values generated by randomly selecting values from a vector
>>   read from a file and stored in memory
>> * values generated by replaying the values in the input file repeatedly
>> Right now, I have a RandomData class that generates all of the 
>> parametric stuff, an EmpiricalDistribution class that does density 
>> estimation from an input file and a ValueServer that uses the two of 
>> these to support data generation in all of the modes described above. 
>> So what I am thinking about is combining them all into one.  The 
>> non-parametric data generation modes are really useful in simulation, 
>> since you can feed them with observed data.  Thoughts?
>> Phil
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail:
>> For additional commands, e-mail:
> ---------------------------------------------------------------------
> To unsubscribe, e-mail:
> For additional commands, e-mail:

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message