SDMM - Simultaneous-direction method of multipliers algorithm

Program code:

function [sol, info] = sdmm(F, param)
%SDMM Simultaneous-direction method of multipliers algorithm
%   Usage: sol = sdmm(F,param);
%          sol = sdmm(F);
%          [sol,info] = sdmm(...);
%
%   Input parameters:
%         F     : Array of function to minimize
%         param : Optional parameter
%   Output parameters:
%         sol   : Solution
%         info : Structure summarizing informations at convergence
%
%   SDMM, from simultaneous-direction method of multipliers solves:
% 
%      sol = argmin sum(f_i( L_i x))
%
%   where x belong to R^N, L_i are linear operators and x_i are the 
%   minimization variables.
%
%   F is a cellarray of structure representing the functions. In the
%   function F{i}, there have to be: 
%
%    F{i}.eval(x_i) : an operator to evaluate the function
%    F{i}.prox(x_i, gamma) : an operator to evaluate the prox of the function
%    F{i}.x0 : vector of initial value 
%
%   Optionally you can define
%
%    F{i}.L  : linear operator, matrix or operator (default identity)
%    F{i}.Lt : adjoint of linear operator, matrix or operator (default identity)
%
%   param a Matlab structure containing solver paremeters. See the
%   function SOLVEP for more information. Additionally it contains those
%   aditional fields:  
%
%    param.tol : is stop criterion for the loop. The algorithm stops if
%
%         max_i ||  y_i(t) - y_i(t-1) ||  / ||y(t) ||< tol,
%      
%     where  y_i(t) are the dual variable of function i at itertion t*
%     by default, tol=10e-4.
%
%       Warning! This stopping criterion is different from other solver!
%
%    param.Qinv : Inverted Q matrix. Qinv=Q^{-1} with:
%
%        Q = sum_i(L_i^t( L_i x))
%
%     By default, Qinv is the identity matrix divided by the number of
%     functions.
%
%     This parameter can be given in a matrix form or in a linear operator
%     form.
%
%   See also: solvep admm forward_backward douglas_rachford
%
%   Demos:  demo_sdmm
%
%   References:
%     P. Combettes and J. Pesquet. Proximal splitting methods in signal
%     processing. Fixed-Point Algorithms for Inverse Problems in Science and
%     Engineering, pages 185--212, 2011.
%     
%     P. Combettes and J. Pesquet. A douglas--rachford splitting approach to
%     nonsmooth convex variational signal recovery. Selected Topics in Signal
%     Processing, IEEE Journal of, 1(4):564--574, 2007.
%     
%
%   Url: https://epfl-lts2.github.io/unlocbox-html/doc/solver/sdmm.html

% Copyright (C) 2012-2016 Nathanael Perraudin.
% This file is part of UNLOCBOX version 1.7.4
%
% This program is free software: you can redistribute it and/or modify
% it under the terms of the GNU General Public License as published by
% the Free Software Foundation, either version 3 of the License, or
% (at your option) any later version.
%
% This program is distributed in the hope that it will be useful,
% but WITHOUT ANY WARRANTY; without even the implied warranty of
% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
% GNU General Public License for more details.
%
% You should have received a copy of the GNU General Public License
% along with this program.  If not, see <http://www.gnu.org/licenses/>.


% Author:  Nathanael Perraudin
% Date: fev 23 2012
%

param.algo = 'SDMM';
[sol, info] = solvep(0,F,param);

end






% function [sol, info,objective] = sdmm(F, param)
% %SDMM Simultaneous-direction method of multipliers algorithm
% %   Usage: sol = sdmm(F,param);
% %          sol = sdmm(F);
% %          [sol,info,objective] = sdmm(...);
% %
% %   Input parameters:
% %         F     : Array of function to minimize
% %         param : Optional parameter
% %   Output parameters:
% %         sol   : Solution
% %         info : Structure summarizing informations at convergence
% %         objective: vector (evaluation of the objectiv function each iteration)
% %
% %   `sdmm`, from simultaneous-direction method of multipliers solves:
% % 
% %   .. sol = argmin sum(f_i( L_i x))
% %
% %   .. math::  sol = \min_x \sum_i f_i( L_i x)
% %
% %   where $x$ belong to $R^N$, $L_i$ are linear operators and $x_i$ are the 
% %   minimization variables.
% %
% %   * *F* is a cellarray of structure representing the functions.
% %
% %     In the function `F{i}`, there have to be:
% %
% %     * `F{i}.eval(x_i)` : an operator to evaluate the function
% %     * `F{i}.prox(x_i, gamma)` : an operator to evaluate the prox of the function
% %     * `F{i}.x0` : vector of initial value 
% %
% %     Optionally you can define
% %
% %     * `F{i}.L`  : linear operator, matrix or operator (default identity)
% %     * `F{i}.Lt` : adjoint of linear operator, matrix or operator (default identity)
% %
% %
% %   * *param* is a Matlab structure containing the following fields:
% %
% %     General parameters:
% %
% %     * *param.tol* : is stop criterion for the loop. The algorithm stops if
% %
% %       ..  max_i ||  y_i(t) - y_i(t-1) ||  / ||y(t) ||< tol,
% %      
% %       .. math:: \max_i \frac{ \| y_i(t) - y_i(t-1)\| }{ \|y_i(t)\|} < tol,
% %
% %       where  $y_i(t)$ are the dual variable of function *i* at itertion *t*
% %       by default, `tol=10e-4`.
% %
% %       Warning! This stopping criterion is different from other solver!
% %
% %     * *param.maxit* : is the maximum number of iteration. By default, it is 200.
% % 
% %     * *param.verbose* : 0 no log, 1 print main steps, 2 print all steps.
% %
% %     * *param.gamma* : convergence parameter (default 1)
% %
% %     * *param.abs_tol* : If activated, this stopping critterion is the
% %       objectiv function smaller than *param.tol*. By default 0.
% %
% %     * *param.Qinv* : Inverted Q matrix. $Qinv=Q^{-1}$ with:
% %
% %       .. Q = sum_i(L_i^t( L_i x))
% %
% %       .. math::  Q = \sum_i L_i^T ( L_i x)
% %
% %       By default, Qinv is the identity matrix divided by the number of
% %       functions.
% %
% %       This parameter can be given in a matrix form or in a linear operator
% %       form.
% %
% %   info is a Matlab structure containing the following fields:
% %
% %   * *info.algo* : Algorithm used
% %
% %   * *info.iter* : Number of iteration
% %
% %   * *info.time* : Time of exectution of the function in sec.
% %
% %   * *info.final_eval* : Final evaluation of the objectivs functions
% %
% %   * *info.crit* : Stopping critterion used 
% %
% %   * *info.rel_norm* : Relative norm at convergence 
% %
% %
% %   See also:  admm forward_backward douglas_rachford
% %
% %   Demos:  demo_sdmm
% %
% %   References: combettes2007douglas combettes2011proximal
% 
% 
% % Author:  Nathanael Perraudin
% % Date: fev 23 2012
% %
% 
% % Start the time counter
% t1 = tic;
% 
% % test the evaluate function
% [F] = test_eval(F);
% 
% % number of function
%     m = size(F,2);
% 
% % Optional input arguments
% if nargin<2, param=struct; end
% 
% if ~isfield(param, 'tol'), param.tol=10e-4 ; end
% if ~isfield(param, 'maxit'), param.maxit=200; end
% if ~isfield(param, 'verbose'), param.verbose=1 ; end
% if ~isfield(param, 'lambda'), param.lambda=1 ; end
% if ~isfield(param, 'gamma'), param.gamma=1 ; end
% if ~isfield(param, 'Qinv'), param.Qinv=@(x) x./m; end
% 
% if isa(param.Qinv,'numeric')
%     param.QinvOp= @(x) param.Qinv*x;
% else
% 	param.QinvOp= param.Qinv;
% end
% 
% for i=1:m
%     if ~isfield(F{i}, 'L'), F{i}.L=@(x) x; end
%     if ~isfield(F{i}, 'Lt'), F{i}.Lt=@(x) x; end
%     if size(F{i}.L,1)==0, F{i}.L=eye(length(F{i}.x0)) ; end
%     if size(F{i}.x0,2)>size(F{i}.x0,1), F{i}.x0=F{i}.x0'; end
%     F{i}.y_n=F{i}.x0;
%     F{i}.z_n=F{i}.x0;
%     F{i}.y_old=F{i}.x0;
%     % Check how the matrix is given
%     if isa(F{i}.L,'numeric')
%         F{i}.OpL= @(x) F{i}.L*x;
%     else
%         F{i}.OpL= F{i}.L;
%     end
%     
%     if isa(F{i}.Lt,'numeric')
%         F{i}.OpLt= @(x) F{i}.Lt*x;
%     else
%         F{i}.OpLt= F{i}.Lt;
%     end
% end
% 
% % Initialization
% x_n=F{1}.OpLt((F{1}.y_n-F{1}.z_n));
% for i=2:m
%     x_n=x_n+F{i}.OpLt((F{i}.y_n-F{i}.z_n));
% end
% x_n=param.QinvOp(x_n);
% 
% [~,~,prev_norm,iter,objective,~] = convergence_test(gen_norm(x_n,F));
% 
% % Main loop
% while 1
%     
%     %
%     if param.verbose >= 1
%         fprintf('Iteration %i:\n', iter);
%     end
%     
%     
%     reldual=0;
%     % Algorithm
%     for i=1:m
%         s_n=F{i}.OpL(x_n);
%         F{i}.y_old=F{i}.y_n; % for convergence criterion
%         F{i}.y_n=F{i}.prox(s_n+F{i}.z_n,param.gamma);
%         F{i}.z_n=F{i}.z_n+s_n-F{i}.y_n ;% updates
%         temp=norm(F{i}.y_old-F{i}.y_n)/norm(F{i}.y_n);
%         if temp > reldual
%             reldual=temp;
%         end
%     end
%     x_n=F{1}.OpLt((F{1}.y_n-F{1}.z_n));
%     for i=2:m
%         x_n=x_n+F{i}.OpLt((F{i}.y_n-F{i}.z_n));
%     end
%     x_n=param.QinvOp(x_n);
%     sol=x_n; 
% 
%     
%     % Global stopping criterion
%     curr_norm = gen_norm(sol,F);  
%     [~,rel_norm,prev_norm,iter,objective,crit] = convergence_test(...
%             curr_norm,prev_norm,iter,objective,param);
%     [sol,param] = post_process(sol, iter, curr_norm, prev_norm, objective, param);
%     if reldual<param.tol
%         break;
%     end
%     if iter>=param.maxit
%         break;
%     end
%     if param.verbose >= 1
%         fprintf([' ||f|| = %e, rel_norm = %e\n Maximum relative ' ...
%             'distance of dual variable: %e\n'], ...
%             curr_norm, rel_norm, reldual);
%     end
% 
%     
% 
%     
% end
% 
% % Log
% if param.verbose>=1
%     %Print norm
%     fprintf('\n Solution found:\n');
%     fprintf(' Final relative norm: %e\n', rel_norm );
%     
%     
%     % Stopping criterion
%     fprintf(' %i iterations\n', iter);
%     fprintf(' Stopping criterion: %s \n\n', crit);
%     
% end
% 
% info.algo=mfilename;
% info.iter=iter;
% info.final_eval=curr_norm;
% info.crit=crit;
% info.time=toc(t1);
% info.rel_norm=rel_norm;
% 
% end
% 
% function n=gen_norm(x_n,F)
% % number of function
%     m = size(F,2);
%     n=0;
%     for i=1:m
%        n=n+F{i}.eval(F{i}.OpL(x_n)); 
%     end
% end