Tue Jan 4 06:50:00 GMT 2000
[ this is a repost of my original e-mail (as is) appended with my new reply ]
Hello. GSL looks something what I have wanted to do for a long time
(without better knowledge of the scientific algorithms, however).
Quick question: can GSL find a sampled function f with approximate
derivatives fd (computed each time from f) which satisfy some
equation S(fd) == 0 with boundary conditions f = 0 and -m < f[i] < m?
I have only experience in using conjugate gradient method from Numerical
Recipes without any boundary conditions...
I did put together a simple list of ideas what I have had earlier, just
in case it has any help for you. By the way, your current TODO looks quite
I'm involved myself in another GNU project (audio software) but because
I need to find out how to find the above mentioned function, I may give
a try and write at least some additional tutorials.
I think an educative documentation would be needed. A quick look at
your documentation revealed that there is a place for improvements.
A documentation which would serve people like me, who doesn't even
know what method to use and how, is a must.
Simple idea list is appended below.
Basic idea is to go systematically through various other libraries and
their documentation for listing functions, algorithms, and references to
literature. That can be handled by non-programmers. Various papers or
algorithm descriptions could be copied (scanned in) for people who are doing
implementations. That is perfectly legal due copyright's fair use as far
as the papers are deleted after use.
the whole book is online;
url? something is in the Netlib; at netlib.att.com(?) www.netlib.org(?)
we could read the original papers and reimplement the methods if
TOMS are not in public domain;
their documentation gives references for algorithms used;
they have new F90/C documentation online, and the older
Fortran library documentation only partially;
IEEE DSP library
this Fortran library is available from the net;
From: Mark Galassi <email@example.com>
>Also, we try to stay clear of numerical recipes, since the authors
>aggressively oppose free software, and we want to give them no cause
>to attack us (even if they would be wrong). So it is best to just
>look at the numerical recipes table of contents and implement all the
>categories of numerical s/w that they identify.
I would want to believe we have a right to read their text too as
far as we don't look at their source code. It is perfectly ok to read
about an algorithmic improvement and implement it ourself. This has
been done with books and papers without source codes.
In straight sense it is the source code we cannot copy, not the
Ok, please let me know what you think about going systematically through
various library documentations, journal papers, books, and softwares.
It would be quite boring and systematic work to go through all this
but it would produce a list of all possible algorithms etc. and also
an extensive reference listing.
There are also other free software for scientific computing which could
be checked better. I remember there is even a GNU software for optimization
More information about the Gsl-discuss