Re: [GNU Octave] interval 2.0.0 released
Thank you so much Oliver,
For your full explanation.
On 11/25/16, Oliver Heimlich <oheim@xxxxxxxxx> wrote:
> On 25.11.2016 06:34, Mehran Mazandarani wrote:
>> Dear Oliver,
>> First of all thank you for your effort on preparing such a useful and
>> applicable software.
>> Would you please let me know how this software calculates the
>> difference between two interval numbers e.g. "u" and "v", "u-v"?
>>
>> Best wishes,
>> Mehran
>
> Dear Mehran,
>
> the implementation uses set-based interval arithmetic and the intervals
> are represented by inf-sup boundaries stored in binary64 format. So, I
> guess you know where this is going …
>
> To compute “x - y” the software uses the following algorithm:
>
> l = inf (x) - sup (y)
> u = sup (x) - inf (y)
>
> if isEmpty (x) or isEmpty (y) then
> result is [Empty]
> else
> result is [l, u]
> end
>
> If x and y are decorated intervals, the software does all of the above
> and then computes the minimum decoration of x, y, and newDec
> (intervalPart (x - y)).
>
> You may look at the source code in [1] and [2]. The subtraction results
> in the algorithm are computed using the MPFR library [3] and outward
> rounding of floating point numbers.
>
> Best
> Oliver
>
>
> [1]
> https://sourceforge.net/p/octave/interval/ci/default/tree/inst/@infsup/minus.m
> [2]
> https://sourceforge.net/p/octave/interval/ci/default/tree/inst/@infsupdec/minus.m
> [3] http://www.mpfr.org/
>
>
--
Mehran Mazandarani
Department of Electrical Engineering
Ferdowsi University of Mashhad, Mashhad, Iran.
homepage:http://mehran.mazandarani.fumblog.um.ac.ir/
IEEE Member, me.mazandarani@xxxxxxxx