By "reliable, high-quality, modern", I mean an order of magnitude
better than Numerical Recipes, of course.
Any suggestions appreciated, but I am not optimistic :-(
Regards,
Nick Maclaren.
I'm curious...
What do you think are the books containing
a) reliable
b) reliable, high-quality
c) reliable, high-quality, modern
algorithms for numerical tasks?
I'm not trying to bother you, I'm really curious.
Thanks in advance,
Fernando.
There are a fair number of those that hit the first two: one classic
is Wilkinson and Reinsch "The Algebraic Eigenvalue Problem".
Or even Acton "Numerical Methods that Work".
Regards,
Nick Maclaren.
I love Acton, but he's hardly got half a dozen algorithms in the whole
book! It's much more of a lore book than an algorithm book.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net
Thanks. I use "Matrix Computations" by Golub and van Loan as a
reference in linear algebra, it includes symmetric and unsymmetric
Eigenvalue problems.
Fernando.
> This question has been asked before, but I have been asked yet again
> to recommend reliable books on high-quality, modern algorithms for
> numerical tasks. Ideally, fairly general ones, but ones for specific
> fields would also be useful.
I can recommend
Goloub & van Loan (1996) Matrix Computations.
Johns Hopkins University Press.
for anything that deals with linear algebra.
Algorithms from pseudo-code can easily be implemented in Fortran 95.
It has references to LAPACK and BLAS routines, and some material on
MPI-style parallel algorithms.
Sturla
I would recommend "Introduction to Numerical Analysis" by
F.B. Hildebrand. The only caveat is that there is not an
ort of "computer code" in the book. It contains just math
and algorithms.
--
steve
Collected Algorithms From ACM
> > Nick Maclaren.- Hide quoted text -
>
> - Show quoted text -
I was going to mention Acton until I scanned the replies and saw this
mention already made.
Of course, you left out the part of the title that is hard to cite well
in a purely textual form: the "usually" that is highlighted on the cover
by the embossing of slected letters in the rest of the title.
I suppose you are right that it is not particularly modern. But it is
still worth reading.
--
Richard Maine | Good judgment comes from experience;
email: last name at domain . net | experience comes from bad judgment.
domain: summertriangle | -- Mark Twain
I don't believe that you can fit into a book the breadth of
coverage of NR, and at significantly more depth. (Quality is
likely not linear with depth (pages), so maybe only two or
three times the pages.)
As I remember it, NR supplies references for more depth on each
specific topic (more or less chapter). If you go down the list
and select about 10 of them, (you probably recognize the names),
you should have a high quality library of numerical methods books.
Though it seems usual for the deeper books to be more theory
and less practice (sample code).
-- glen
Why prohibit type checking? Fortran 90 code with modules or interfaces
has it.
The wording is slightly awkward. I had to read it a second time myself
because it didn't seem to make sense the first time. But if I read it
correctly on the second pass, I think you missed the double negative;
that is, it should NOT require no type checking.
I inferred the wording was specifically aimed at rejecting the texts
written with the Matlab's of the world in mind...
--
I apologise for being gratuitously obfuscatory!
As Richard has pointed out before, I tend to think and write in
double negatives a bit much for clear communication. I was trying
to say that the books should be compatible with implementation in
Matlab or straightforward Fortran.
There are occasional books that assume an unchecked language, or
one with dynamic typing, and they are murder to implement in
languages with strong, usually static, typing. The same is true
of the ones that assume that code can be created and executed
dynamically.
Regards,
Nick Maclaren.
There aren't any.
I quite like
"Accuracy and Stability of Numerical Algorithms"
by Nicholas Higham
(I have the second edition, published by SIAM in 2002). Of course its scope
is somewhat limited, but it has a lot of rather up-to-date references to
modern methods in addition to extensive error analysis etc. It is unfortunately
also quite expensive (> 55 EUR).
Regards
Reinhold
My problem here, which other people will recognise, is that far too
many students nowadays don't want understanding - they want recipes.
This does not fit well with numerical computation of real problems,
especially the non-trival ones :-(
Sparse matrices are again not one my areas, but there should be a
new version of Duff, Erisman and Reid "Direct Methods for Sparse
Matrices" shortly. I think that we can guarantee that the book
will be Fortran-friendly :-)
Regards,
Nick Maclaren.
> This question has been asked before, but I have been asked yet again
> to recommend reliable books on high-quality, modern algorithms for
> numerical tasks. Ideally, fairly general ones, but ones for specific
> fields would also be useful. They should also be implementable in
> Fortran - i.e. should NOT require dynamic compilation, no type
> checking and other such aberrations.
Ronald Thisted. Elements of Statistical Computing: Numerical Computation
ISBN-13: 978-0412013713.
http://galton.uchicago.edu/~thisted/Distribute/Errata.pdf