The slp_sm (sparse linear prediction with splitting methods) solves 
the sparse linear prediction problem 

   minimize   ||x-Xa||_1 + gamma ||a||_1
      a

where X is a convolution matrix formed from the signal x. 

The considered methods for solving the above problem are
  1: Douglas-Rachford (DR)
  2: Alternating direction method of multipliers (ADMM)

Provided is both a Matlab and a C++ implementation.
    mlib/utilities: Matlab implementation
    src: C++ implementation

This software is used in the paper

   Fast Algorithms for High-Order Sparse Linear Prediction
   with Applications to Speech Processing

   Tobias L. Jensen, Daniele Giacobello, Toon van Waterschoot
   and Mads G. Christensen
   
   Speech Communication, 2016

   Please give reference if you use this software.


= MATLAB:
=========

To use the matlab implementation, it is necessary to compile two c-files. To do 
this first run "make" in the root directory of this software. If you do not have a c-compiler, follow the instructions on the screen.

Then add /mlib and /utilities to your path.

An example of usage is given in /example/example_dr_admm_slp.m

Location:
   /mlib and /utilities

Requirements:
    + A c-compiler

The following libraries are necessary for running the tests:

  - mlib/test.m and utilities/test.m
    + Matlabs unittest framework. Available from release 2013a 
    + cvx: http://cvxr.com/cvx/


= C++ implementation:
=====================

Location:
   /src

The following libraries are are necessary for running the programs:
    + fftw3 library (fftw.org)
    + BLAS and LAPACK libraries
    / optional Math Kernel Library (MKL)


The following libraries are necessary for running the tests in src/unittest:
    + Googletest,  https://code.google.com/p/googletest/
    + fftw3 library (fftw.org)
    + BLAS and LAPACK libraries
    / optional Math Kernel Library (MKL)

The C++ source code is implemented to benefit from Math Kernel Library
If you want to make use of this, compile with -DMKL and add the necessary
libraries to the link command. Otherwise, a link command along the lines 
-lblas -llapack -lfftw3 -lfftw3f should do. See src/Makefile