Hmmm. I won't dispute the fact that the methods described in
NR are aged or that the texts are aimed a numerical-analysis
neophytes. Both of these seem to be somewhat accurate. However,
this does not mean that the books are useless. I took a numerical
methods course at the University of Florida as an undergraduate and
while we used another textbook at the time I think the NR series of
books with some guidance and explanations by a competent instructor
would be excellent as a senior undergraduate introduction to numerical
methods programming. The other book we used (I think it was by
Chapra and Canale) was way too simplistic and superficial. At the
time I was taking this course at UF chemical engineering I had purchased
the first edition of "Numerical Recipes in C" published in 1988.
I admit the book seemed a bit terse then but in retrospect I now see that
had our instructor used this book or one of the Fortran analogues along
with some notes and lectures explaining the methods and techniques (NR
is overly verbose and misleading sometimes) we would have been much better
off. I got an A in the course but I never really use the book in the course
for reference. When I want to find out more about a particular technique I
use the NR book in C I have. In fact it was this book that allowed me to
write a very nice piping network solver as an undergraduate at UF. It has
also helped me out on several other small but useful numerical projects.
Clearly I am not a numerical-analyst. I simply want a reference book on
some of the older and more common techniques in the field so that I can
implement them to solve my own problems. I would rather do that than use
a canned program like mathematica or matlab. By the same token though, if I
were to do numerical analysis for a living I would certainly not be using
NR. More likely I would be using the most modern techniques currently
available
in the scientific literature and notes from lectures or symposiums. My
point
is this. If you are a professional numerical analyst you probably have a
more thorough understanding of the issues involved in numerical analysis
than
the average engineer or scientist and as such would likely find the NR books
elementary and perhaps somewhat insulting. In that case you will more
likely by making use of other references. On the other hand the average
engineer or scientist like myself does use the NR book on occasion and
keeps a copy of it on the shelf for reference when needed. If the numerical
problem I am trying to solve becomes so intractable that the techniques
presented in NR are of no use then I go see a numerical analyst for advice.
:-)
Juan Casero
[log in to unmask]
-----Original Message-----
From: [log in to unmask]
[mailto:[log in to unmask]]On Behalf Of Van Snyder
Sent: Friday, September 29, 2000 2:24 PM
To: [log in to unmask]
Subject: Runge-Kutta (again)
Sabino Chavez-Cerda has restated a common observation:
> Although some NR subroutines may have some problems, many others
> are well and accurate.
The authors of NR have a charming and lucid writing style. It gives
one the impression that the entire topic of numerical computation is
easy. As one reads NR, one also gains the impression that both the
reader and the authors understand the material completely.
Unfortunately, none of this is true.
It is precisely because of the conspiracy of this seductive writing
style, and the mixed quality of the work it so pleasingly exposes,
that a neophyte, at whom the book is consciously aimed, is likely to
be led astray. An expert can separate the wheat from the chaff, but
then an expert wouldn't be using NR for a textbook.
This will continue to be true so long as the authors of NR pursue
overly-simple solutions to complex problems, or continue to describe
methods that have been obsolete for decades. My own experience, as
a result of exchange of several letters with one of the authors of
NR (concerning MEDFIT) is that this situation is unlikely to change.
The upshot of that exchange was that I became convinced that that
author never understood the faulty mathematical foundation of the
algorithm, even after four careful explanations.
Chavez-Cerda's observation, and other factors described above, are
reasons both to avoid and to admire NR.
--
What fraction of Americans believe | Van Snyder
Wrestling is real and NASA is fake? | [log in to unmask]
Any alleged opinions are my own and have not been approved or disapproved
by JPL, CalTech, NASA, Dan Goldin, Bill Clinton, the Pope, or anybody else.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|