SI-AMP

SI-AMP algorithm

Introduction

The code implements the proposed algorithm GENP-AMP in the paper “Approximate Message Passing-based Compressed Sensing Reconstruction with Generalized Elastic Net Prior” (http://arxiv.org/abs/1311.0576) and reproduces the experimental parts.

Requirements

1. Mathworks MATLAB release 2009b or later

2. The CVX software package, (available at http://cvxr.com/cvx/ )

3. The GAMP software package, (available from from Sourceforge at http://sourceforge.net/projects/gampmatlab/files/), installed and included in MATLAB's path.

All related software packages are included in the document. If the readers want to get the latest version, please download from the webpages above.

Installation instructions:

1. Install the CVX package from http://web.cvxr.com/cvx/doc/ and include the folders in MATLAB’s path.

2. Install the GAMP software packages from http://sourceforge.net/projects/gampmatlab/files/ and include the folders in MATLAB’s path

3. The multi-view dataset can be downloaded from http://www.fujii.nuee.nagoya-u.ac.jp/multiview-data/ .

Usage

How to use SI-OWLQN:

• Save the data into .mtx format ,e.g.,

  [ err1 ] = mmwrite( 'A_matrix.mtx',A); % %

  [ err2 ] = mmwrite( 'y_matrix.mtx',Y); %

  [ err3 ] = mmwrite( 'SI_matrix.mtx',x_SI); %

  [ err4 ] = mmwrite('x_matrix.mtx',x);

  The mmwrite and mmread programme have been included in sub-folder "solver".

• cd to comparison algorithms\SI-OWLQN

• The input format should be like this:

  SI-OWLQN x_matrix.mtx A_matrix.mtx y_matrix.mtx SI_matrix.mtx lambda output.mtx –l2weight tau