help-octave
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Zero or not?


From: Przemek Klosowski
Subject: Re: Zero or not?
Date: Mon, 23 Jul 2007 11:30:18 -0400 (EDT)

   Ok, that's a start. If I calculate the sum of N numbers, what is the  
   maximum 'offset' of the numerical result to mathematically correct  
   result? I'd guess its N*eps, but I'm not sure.

THat's hard to answer, because your maximum error would be N times the
error of each component, and so it depends what is the error of each
component. Given decent numerical algorithms, you'd expect it would be
on the order of a LSB, i.e. eps, but 'It Depends' (TM). By the way, if
the indivudual component errors are independent, the sum's expected
error is sqrt(N)*eachComponentError.

This is not really an octave issue, but rather a general numerical
analysis problem, so if it is important to your application you should
read the Goldberg paper and numerical analysis texts relevant to your
algorithm.

        greetings

        p


reply via email to

[Prev in Thread] Current Thread [Next in Thread]