Copyright 2012 Erlend Aune

THIS LIBRARY IS NO LONGER MAINTAINED.
It will be integrated into SHOGUN: 
(https://github.com/shogun-toolbox/shogun)
and future progress will be found there.

The Krylov statistics library (KRYLSTAT) is free C++ software
under the LGPL license. It is designed to facilitate sampling
from high dimensional Gaussian distributions using rational
approximations and Krylov methods and computing log-determinants
using the same methods with the addition of graph colouring.


The code depends on ARPREC for computing high-precision Jacobi
elliptic functions. The libRatApp.a-file includes this, but
the header file for arprec is needed. The library can be found 
on http://crd.lbl.gov/~dhbailey/mpdist/. 

It depends on Eigen (http://eigen.tuxfamily.org/index.php)
for blas-type functions and sparse matrix-vector products.

On ColPack  (http://www.cscapes.org/coloringpage/software.htm) 
for graph colouring.

On boost (http://www.boost.org/) for computing IID normal
samples.

and on cusp (http://code.google.com/p/cusp-library/) for 
GPU implementations.

Additionally, OpenMP is required for parallel computations.


Citing this software may be done by citing one or both of 
the following bibtex entries:

@ARTICLE{aunsimp_par_est,
  author = {Erlend Aune and Daniel P. Simpson and Jo Eidsvik},
  title = {Parameter estimation in high dimensional Gaussian distributions},
  journal = {Statistics and Computing},
  year = {2013},
  volume = {To appear},
  pages = {NA},
}

@ARTICLE{aun_samp_stco_2012,
  author = {Erlend Aune and Jo Eidsvik and Yvo Pokern},
  title = {terative Numerical Methods for Sampling from High Dimensional Gaussian Distributions},
  journal = {Statistics and Computing},
  year = {2012},
  volume = {To appear},
  pages = {NA}
}