bugGNU Scientific Library - Bugs: bug #30510, problems with hyperg_U(a,b,x) for...

 
 

You are not allowed to post comments on this tracker with your current authentication level.

bug #30510: problems with hyperg_U(a,b,x) for x<0

Submitter:  -Deleted Account- <bjg>
Submitted:  Wed 21 Jul 2010 08:17:45 PM UTC
   
 
Category:  Runtime error Severity:  3 - Normal
Operating System:  Status:  None
Assigned to:  None Open/Closed:  Open
Release:  1.14

Wed 21 Jul 2010 08:17:45 PM UTC, original submission:  

Reply-To: -email is unavailable-
From: Raymond Rogers <raymond.rogers72@gmail.com>
To: -email is unavailable-
Subject: Re: [Bug-gsl] hyperg_U(a,b,x) Questions about x<0 and values
Date: Thu, 08 Jul 2010 11:49:15 -0500

Brian Gough <bjg@gnu.org>
Subject: Re: [Bug-gsl] hyperg_U(a,b,x) Questions about x<0 and values
        of a
To: -email is unavailable-
Cc: -email is unavailable-
Message-ID: <43wrt61fkt.wl%bjg@gnu.org>
Content-Type: text/plain; charset=US-ASCII

At Wed, 07 Jul 2010 10:14:34 -0500,                                                                                           
Raymond Rogers wrote:                                                                                                         

> >                                                                                                                           
> > 1) I was unable to find the valid domain of  the argument a when x<0.                                                     
> > Experimenting yields what seem to be erratic results.  Apparently                                                         
> > correct answers occur when {x<0&a<0& a  integer}.  References would be                                                    
> > sufficient.   Unfortunately {x<0,a<0} is exactly the wrong range for my                                                   
> > problem; but the recursion relations can be used to stretch to a>0.  If                                                   
> > I can find a range of correct operation for the domain of "a" of width >1.                                                
>                                                                                                                             


| Brian Gough                                                                                                                 
| Thanks for the email.  There are some comments about the domain for                                                         
| the hyperg_U_negx function in specfunc/hyperg_U.c -- do they help?                                                          

They explain some things, but I believe the section

  if (b_int && b >= 2 && !(a_int && a <= (b - 2))){}
  else {}

is implemented incorrectly; and probably the preceding section as well.  Some restructuring of the code would make things clea\
rer; but things like that should probably done in a different forum: email, blog, etc...
I think the switches might be wrong.  In any case it seems that b=1 has a hole.  Is there a source for this code?

Note: the new NIST Mathematical handbook might have better algorithms.  I am certainly no expert on implementing mathematical \
functions (except for finding ways to make them fail).

Ray

Reply-To: -email is unavailable-
From: Raymond Rogers <raymond.rogers72@gmail.com>
To: -email is unavailable-
Subject: [Bug-gsl]      Re: hyperg_U(a, b, x) Questions about x<0 and values of a,
Date: Sun, 11 Jul 2010 14:43:43 -0500

hyperg_U basically fails with b=1, a non-integer; because
gsl_sf_poch_e(1+a-b,-a,&r1); is throwing a domain error when given
gamma(0)/gamma(a).
Checking on and using b=1 after a-integer is checked is illustrated
below in Octave.   I also put in recursion to evaluate b>=2.
I checked the b=1 expression against Maple; for a few values x<0,a<0,b=1
and x<0,a<0,b>=2 integer.
--------------
Unfortunately the routine in Octave to call hyperg_U  is only set up for
real returns, which was okay for versions <1.14 .  Sad to say I am the
one who implemented the hyperg_U interface, and will probably have to go
back :-( .  Integrating these functions into Octave was not pleasant;
but perhaps somebody made it easier.  I did translate the active parts
of hyperg_U into octave though; so it can be used in that way.

Ray

#
#
# Test function to evaluate b=1 for gsl 1.14 hyperg_U x<0
#
function anss=hyperg_U_negx_1(a,b,x)
    int_a=(floor(a)==a);
    int_b=(floor(b)==b);
#neg, int, a is already taken care of so use it
    if (int_a && a<=0)
        anss=hyperg_U(a,b,x);
    elseif (int_b && (b==1))
#from the new NIST DLMF 13.2.41

anss=gamma(1-a)*exp(x)*(hyperg_U(1-a,1,-x)/gamma(a)-hyperg_1F1(1-a,1,-x)*exp((1-a)*pi*I));
    elseif (b>=2)
#DLMF 13.3.10
        anss=((b-1-a)*hyperg_U_negx_1(a,b-1,x) +
hyperg_U_negx_1(a-1,b-1,x))/x;
    else
        anss=hyperg_U(a,b,x);
    endif
#
endfunction




-Deleted Account- <bjg>

 

(Note: upload size limit is set to 16384 kB, after insertion of the required escape characters.)

No files currently attached

 

Depends on the following items: None found

Items that depend on this one: None found

 

Carbon-Copy List
  • -email is unavailable- added by bjg
  •  

    Follows 1 latest change.

    Date Changed by Updated Field Previous Value => Replaced by
    2010-07-21 bjg Carbon-Copy- Added -email is unavailable-

    Back to the top

    Powered by Savane 3.14-8eb0.
    Corresponding source code