[ANN] Sci 1.0-beta4, LuaJIT Universal Package 2.0.3, others

From: Stefano <phd.st.p@xxxxxxxxx>
To: luajit@xxxxxxxxxxxxx
Date: Sun, 23 Mar 2014 16:44:52 +0000
Hi,

having managed to write the documentation for it I just released a new
version of Sci - a library for general purpose scientific computing
(MIT licensed). It's composed of the following modules:

sci.math: special mathematical functions
sci.diff: automatic differentiation
sci.alg: vector and matrix algebra
sci.quad: quadrature algorithms
sci.root: root-finding algorithms
sci.fmin: function minimization algorithms
sci.fmax: function maximization algorithms
sci.prng: pseudo random number generators
sci.qrng: quasi random number generators
sci.stat: statistical functions
sci.dist: statistical distributions
sci.mcmc: MCMC algorithms

At the same time I updated all my other libraries (Xsys, Time,
LJSQLite3, Rclient) and released a new version of the LuaJIT Universal
Package for LuaJIT 2.0.3 which as usual supports Windows, Linux, OS X
and bundles a number of system libraries. Thanks to the luapower
developer(s) who inspired some simplifying changes.

To avoid confusion I numbered all libraries as 1.0-beta4. As usual,
everything is available for download at:

http://www.scilua.org/

I would suggest to just decompress everything into a new directory and
move there your personal libraries / work.

As it proved to be a popular request I have created Git repositories
for all the libraries:

https://github.com/stepelu

It's the first time I use Github, so if you spot something which is
clearly wrong please let me know and I'll fix it.

As I know that there are a number of users around who tired my
libraries I would really appreciate some feedback / features requests.
In particular, the sci.alg module (definitely the most complex one) is
the outcome of a number of re-writes. I added a short implementation
documentation in the file in case someone is interested to have a
peek.

Expect the following features to appear in the next relase(s):
+ faster forward-mode automatic differentiation: I have implemented a
prototype which requires a single traversal of the function, around
30-40% faster in typical multivariate scenarios. Then backward-mode
automatic differentiation...
+ multivariate statistical distributions: uniform, normal, student,
dirichlet, copula
+ BFGS optimizer
+ higher-level BLAS/Lapack-based matrix algebra routines

I'll be traveling in the next month hence my replies could be delayed.


Stefano
[ANN] Sci 1.0-beta4, LuaJIT Universal Package 2.0.3, others

Other related posts:
