Description:
~~~~~~~~~~~

This file contains the R source code of the model considered in the paper
"Learning Feature Selection Dependencies in Multi-task Learning" by D. 
Hernandez-Lobato and J. M. Hernandez-Lobato.

The code is implemented in R but makes calls to a C implementation of
a numerical quadrature function for fast implementation. Thus, it is required 
first to compile the C code. This is done from a Linux prompt by typing:

R CMD SHLIB refinePriorTermsHSFast.c

Two libraries are required: gls and gslcbla these can be installed
with the command: 

sudo apt-get libgsl0-dev

Once the code has been compiled one only has to run the R code contained
in example.R.

The file  horseShoeEP_4Q.R contains the actual implementation of the method
described in the paper.

To run the example one typically introduces the command:

R --no-save < example.R

It takes about 1h running on a i5 comptuer. It requires the R package doMC
to be installed.

