[Top][All Lists]

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

[Bug-gsl] [bug #45053] gsl_min_find_bracket is most likely incorrectly i

From: Chris Mansley
Subject: [Bug-gsl] [bug #45053] gsl_min_find_bracket is most likely incorrectly implemented
Date: Thu, 07 May 2015 23:26:22 +0000
User-agent: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Ubuntu Chromium/37.0.2062.120 Chrome/37.0.2062.120 Safari/537.36


                 Summary: gsl_min_find_bracket is most likely incorrectly
                 Project: GNU Scientific Library
            Submitted by: cmansley
            Submitted on: Thu 07 May 2015 11:26:20 PM GMT
                Category: Runtime error
                Severity: 3 - Normal
        Operating System: ubuntu 12.04 (kernel 3.8.0-44-generic)
                  Status: None
             Assigned to: None
             Open/Closed: Open
                 Release: 1.15
         Discussion Lock: Any



There is a function for pre-bracketing the minimum for the
gsl_min_fminimizer_* set of functions called gsl_min_find_bracket. This
function can be forced to evaluate the function outside of the initial
supplied bracket.

This is extremely bad as the function may not be defined outside of the
initial range.

The attached program, gsl_bug.c, forces the bracketing to search outside of
the initial range on a parabola, which should be a simple bracketing

The attached patch follows the formulations formulations of bracketing from
Brent's algorithm and the golden area search, by reducing the bracket area at
each step.

In gsl_bug.c, the patched function performs a better bracketing with fewer

I did not exhaustively test this for failure cases, but it seems like it does
a better job fulfilling the specifications.

I would like to hear if this was intended behavior or an actual bug.   


File Attachments:

Date: Thu 07 May 2015 11:26:20 PM GMT  Name: gsl_bug.c  Size: 766B   By:

Date: Thu 07 May 2015 11:26:20 PM GMT  Name: bracketing.patch  Size: 970B  
By: cmansley



Reply to this item at:


  Message sent via/by Savannah

reply via email to

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