[Top][All Lists]

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

[Bug-gsl] Minor but still obvious stylistic concerns in BFGS minimizatio

From: Alan W. Irwin
Subject: [Bug-gsl] Minor but still obvious stylistic concerns in BFGS minimization routines
Date: Fri, 14 Apr 2006 00:15:23 -0700 (PDT)

Version gsl-1.7:

(1) directional_minimize.c in the minimize routine:

double iter = 0;

should be changed to

int iter = 0;

Some of the type double comparisons made later in the routine are compared
to 0 rather than 0.

These changes should speed up the routine (since casts are avoided).
I imagine the effect is minute, and possibly not even measurable so that
is why I classify these issues as merely a stylistic concern.

(2) vector_bfgs.c:

The method is incorrectly classified in a few places as a conjugate gradient
method rather than the appropriate classification which is a quasi-Newton
method.  This incorrect classificiation could cause some misunderstandings
for people unfamiliar with minimization methods.

Here are my suggested changes:

(a) In one comment

conjugate gradient method ==> quasi-Newton method

(b) In two output messages:
conjugate direction ==> direction

(3) This same confusion also occurs in
http://www.network-theory.co.uk/docs/gslref/gsl-ref_467.html when the method
is described as follows:

"This is the vector Broyden-Fletcher-Goldfarb-Shanno (BFGS) conjugate
gradient algorithm. It is a quasi-Newton method which builds up an
approximation to the second derivatives of the function f using the
difference between successive gradient vectors."

I think "conjugate gradient" should be dropped from that first sentence.

These are all extremely minor nit-picks, but nevertheless, it would be nice
to get these issues cleaned up.

Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the Yorick front-end to PLplot (yplot.sf.net); the
Loads of Linux Links project (loll.sf.net); and the Linux Brochure Project

Linux-powered Science

reply via email to

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