Difference between Matlab and Octave in random seeding

[dbateman]

Ben Abbott wrote:
> 
> 
> On Mar 7, 2008, at 6:25 PM, John W. Eaton wrote:
> 
>> On  7-Mar-2008, pbellec wrote:
>>
>> | My understanding is that any difference of behavior between matlab  
>> and octave
>> | should be regarded as a bug,
>>
>> That's false.
>>
>> jwe
> 

I consider differences as bugs, but maybe the bug is in Matlab rather than
Octave :-)



> I have no vested interest in this particular discussion, but do have  
> an interest in consistency with Matlab.
> 
> Does this "feature" cross that line?
> 

Matlab has no rande, randp and randg function. You need the equivalent
function in the statistics toolbox like exprnd. However, Octave having
particularly a good randg function makes all the other random sequences
easier to generate. As matlab doesn't have rande, randp and randg the
implementations of exprnd, etc depend on rand/randn and so getting their
behavior wrt seeding the same as matlab means getting rid or
rande/randg/randp. I'd be against this for reasons of speed and simplicity
of the implementation of the random distributions in the statistics toolbox. 

This will have to be one area where Octave is different to matlab, and in
fact better. Check 

http://www.nabble.com/Low-hanging-fruit---Accelerated-random-distribution-functions-to9103622.html#a9113151

and you'll see Octave has basically the best in class random generators..

D.

-- 
View this message in context: http://www.nabble.com/Difference-between-Matlab-and-Octave-in-random-seeding-tp15508096p15914943.html
Sent from the Octave - Bugs mailing list archive at Nabble.com.