ivan@154: function [A] = generate_matrix(M, N, ensemble, p)
ivan@154: % =========================================================================
ivan@154: %                       Measerement matrix generator
ivan@154: % =========================================================================
ivan@154: % INPUT ARGUMENTS:
ivan@154: % M                         Number of measurements (number of rows).
ivan@154: % N                         Size of sparse vector (number of columns).
ivan@154: % ensemble                  Ensemble type of measurement matrix. Possible
ivan@154: %                           values are:
ivan@154: %                           -'Gaussian': creates a MxN measurement matrix with 
ivan@154: %                           elements drawn from normal distribution N(0,1).%                           
ivan@154: %                           -'Bernoulli': creates a MxN measurement matrix with 
ivan@154: %                           elements drawn from Bernoulli distribution (1/2,1/2).
ivan@154: %                           -'pBernoulli': creates a MxN measurement matrix with 
ivan@154: %                           elements drawn from Bernoulli distribution (p,1-p).
ivan@154: %                           Parameter of Bernoulli distribution.
ivan@154: %                           -'sparseGaussian': creates a MxN sparse measurement 
ivan@154: %                           matrix with elements drawn from normal distribution N(0,1).
ivan@154: % =========================================================================
ivan@154: % OUTPUT ARGUMENTS:
ivan@154: % A                     MxN measurement matrix with normalized columns.
ivan@154: % =========================================================================
ivan@154: % 01/04/2011, by Anastasios Kyrillidis. anastasios.kyrillidis@epfl.ch, EPFL.
ivan@154: % =========================================================================
ivan@154: if nargin < 3
ivan@154:     ensemble = 'Gaussian';
ivan@154: end;
ivan@154: 
ivan@154: if nargin < 4
ivan@154:     p = 0.5;
ivan@154: end;
ivan@154: 
ivan@154: switch ensemble
ivan@154:     case 'Gaussian'
ivan@154:         A = randn(M,N);         % Standard normal distribution                
ivan@154:         for i = 1:N             % Normalize columns
ivan@154:             A(:,i) = A(:,i)/norm(A(:,i));
ivan@154:         end;
ivan@154:     case 'Bernoulli'
ivan@154:         A = (-1).^round(rand(M,N));     % Bernoulli ~ (1/2, 1/2) distribution
ivan@154:         for i = 1:N             % Normalize columns
ivan@154:             A(:,i) = A(:,i)/norm(A(:,i));
ivan@154:         end;
ivan@154:     case 'pBernoulli'
ivan@154:         A = (-1).^(rand(M,N) > p);   % Bernoulli ~ (p, 1-p) distribution     
ivan@154:         for i = 1:N             % Normalize columns
ivan@154:             A(:,i) = A(:,i)/norm(A(:,i));
ivan@154:         end;
ivan@154:     case 'sparseGaussian'
ivan@154:         leftd = 8;
ivan@154:         A = zeros(M,N);
ivan@154:         for i = 1:N
ivan@154:             ind = randperm(M);
ivan@154:             A(ind(1:leftd),i)=1/leftd;
ivan@154:         end
ivan@154:         for i = 1:N             % Normalize columns
ivan@154:             A(:,i) = A(:,i)/norm(A(:,i));
ivan@154:         end;
ivan@154: end;