function x = SparseVector(n, k, ensemble, perm)
% SparseVector: Generates a sparse vector with a specified distribution.
%
%  Usage:
%     x = SparseVector(n, k, ensemble, perm)
%  Inputs:
%    n          vector length
%    k          number of nonzero entries
%    ensemble   string containing name of coefficient distribution
%               'Uniform', 'Gaussian', 'Signs', 'Power'. 
%               Default is 'Uniform'.
%    perm       If =1, the nonzero indices are randomly selected. 
%               Otherwise, the nonzero entries are in indices 1..k (default).
%  Outputs:
%    x          Sparse n vector.
%  Description:
%    This function creates a vector of length n with k nonzero entries, 
%    distributed according to the specified input ensemble. 
%    The following distributions are supported:
%
%      'Uniform' - Entries are distributed uniformly on the unit interval. 
%
%      'Gaussian' - Entries are distributed N(0,1).
%
%      'Signs' - Entries are distributed Bernoulli over the set {-1,1}, 
%                with equal probabilities. 
%
%      'Power' - Entries follow the power law 1/j, j = 1..k
%

if nargin < 4,
    perm = 0;
end
if nargin < 3,
    ensemble = 'Uniform';
end

switch upper(ensemble)
    case 'UNIFORM'
        x = [rand(k,1); zeros(n-k,1)];
        
    case 'SIGNS'
        x = sign(rand(k,1) - 0.5);
        zz = find(x == 0);
        Phi(zz) = ones(size(zz));
        x = [x; zeros(n-k,1)];
        
    case 'GAUSSIAN'
        x = [randn(k,1); zeros(n-k,1)];

    case 'POWER'
        x = [1./[1:k]; zeros(n-k,1)];
end        

if perm
    p = randperm(n);
    x = x(p);
end
%
% Part of SparseLab Version:100
% Created Tuesday March 28, 2006
%