*To*: debian-science@lists.debian.org*Subject*: Re: Welcome to Debian Science!*From*: Andre Lehovich <andrel@yahoo.com>*Date*: Mon, 1 Aug 2005 08:29:27 -0700 (PDT)*Message-id*: <20050801152927.88114.qmail@web52406.mail.yahoo.com>*In-reply-to*: <42ED9001.70102@debian.org>

Helen suggests listing which packages we're using in our research. For me the unifying GUI is emacs. Many of the apps I use have nice emacs modes which integrate the running program into the editor. For plotting, quickie computation, and algorithm prototyping I'm currently using R. Getting up to speed with R is definitely an investment, but I'm finding it is paying off nicely. This web page has some nice plotting examples on it: http://www.stat.auckland.ac.nz/~ihaka/120/ R pros: entire enivronment is oriented around data analysis and visualization. Produces high-quality EPS output. Includes a real programming language, which turns out to be important for more sophisticated plots. Very active user community. R cons: a lot to learn. Lack of good DFSG-compliant documentation. For symbolic computation of derivatives I'm using maxima. I haven't yet tried to get integrals out of it. Subroutine libraries I use are GSL, ATLAS, LAPACK, and FFTW. GSL pros: general-purpose numerical analysis subroutines. I like the pseudo-random number generators. Also has ODE solvers, function minimizers, special functions, and so on. GSL cons: linear algebra support is weak. ATLAS/BLAS/LAPACK are the standard numerical linear algebra libraries. BLAS is a set of simple operations, LAPACK builds on it to provide eigenvalue solvers, SVD, etc. Most supercomputer vendors provide an implementation of the BLAS tuned to their architecture. ATLAS is an implementation of BLAS which automatically tunes itself to run efficiently on your CPU. ATLAS/BLAS/LAPACK pros: The standard. Written by experts in numerical linear algebra. ATLAS/BLAS/LAPACK cons: Oriented towards Fortran, i.e. lacking C prototypes and uses column-major layout in memory. I call it from C anyway. FFTW is a library for discrete Fourier transforms FFTW pros: very fast, gracefully handles the non-power-of-two case, tunes itself to your CPU type. FFTW cons: none. For image import/export I use the PNM image formats, with netpbm and imagemagick to do the conversions. To display large images I use gimp. (I'm looking for good free software to work with tomographic datasets. Everything I know of has DFSG-incompatible dependencies.) I've used VTK to do 3D visualization in the past. For logging into the compute cluster on the other coast I use VNC running down an SSH tunnel. This is *much* faster than using SSH's X11 forwarding capability. For writing I use LaTeX. Emacs has some nice extensions to get near-WYSIWYG editing. The ones I use are preview-latex and whizzytex. --Andre __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com

- Prev by Date:
**Re: Welcome to Debian Science!** - Next by Date:
**Re: debian-science@lists.debian.org has been created** - Previous by thread:
**Re: Welcome to Debian Science!** - Next by thread:
**Re: Welcome to Debian Science!** - Index(es):