dictionary_gen_detailed_data.m

Go to the documentation of this file.

function detailed_data =  dictionary_gen_detailed_data(model,model_data)
% simply random parameter set and snapshot generation
%
% B. Haasdonk 5.3.2015

tic;
sim_data = detailed_simulation(model,model_data);
uh = model.get_dofs_from_sim_data(sim_data);
ndofs = length(uh);
if ~isfield(model,'RB_train_data_mode')
  model.RB_train_data_mode = 'random';
end;

detailed_data  = model_data;

if ~isequal(model.RB_train_data_mode,'offline_greedy')
  %%%% Lagrangian Dictionary for random or equidistant points  
  if isequal(model.RB_train_data_mode,'random') 
    mus = rand_uniform(model.RB_train_size,model.mu_ranges);
  elseif isequal(model.RB_train_data_mode,'equidistant') 
    % equidistant
    mu_dim = length(get_mu(model));
    mus = zeros(mu_dim, model.RB_train_size);
    for i=1:mu_dim
      mu_range = model.mu_ranges{i};
      mus(i,:) = linspace(mu_range(1),mu_range(2),model.RB_train_size);
    end;
  elseif isequal(model.RB_train_data_mode,'points_given') 
    mus = model.Mtrain;
  else 
    error('RB_train_dat_mode unknown');
  end;
  RB = zeros(ndofs,model.RB_train_size);
  for mui = 1:model.RB_train_size;
    model = model.set_mu(model,mus(:,mui));
    sim_data = detailed_simulation(model,model_data);
    RB(:,mui) = model.get_dofs_from_sim_data(sim_data);
    fprintf('.');
  end;
  fprintf('\n');
else 
  %%%%%%%%%%%%%%%%%%%%%%%%% Offline Greedy
  Mtrain = model.Mtrain;
  ntrain = size(Mtrain,2);
  RB = zeros(ndofs,0);
  Delta_max = [];
  tmp_detailed_data = model_data;
  model.decomp_mode = 1; % == components
  [tmp_detailed_data.A_comp, tmp_detailed_data.f_comp] = ...
      model.operators(model,detailed_data);
  tmp_detailed_data.RB = RB;
  mu_indices = [];
  mus = zeros(size(Mtrain,1),0);
  Deltas_max = [];
  
  % take first basis vector arbitrary
  %i0 = 1;
  %model = set_mu(model,Mtrain(:,i0));
  %sim_data = detailed_simulation(model,model_data);
  %RB = [RB,model.get_dofs_from_sim_data(sim_data)];  
  %mus = Mtrain(:,i0);
  %mu_indices = [i0];

  reduced_data =  gen_reduced_data(model,tmp_detailed_data);

  for N = 1:model.RB_stop_Nmax;
    disp(['processing offline greedy extension ',num2str(N)]);
    Delta_max = 0;
    for i=1:ntrain;
      model = set_mu(model,Mtrain(:,i));
      sim_data = rb_simulation(model,reduced_data);
      %    if res.res_norms(end)>res_max
      %      res_max = res.res_norms(end);
      %      end;
      Delta = sim_data.Delta;
      if Delta>Delta_max
        Delta_max = Delta;
        i_max = i;
      end;
    end;
    % take snapshot for worst parameter into basis
    disp(['maximum Delta ',num2str(Delta_max)]);
    if N>1 & Delta_max > Deltas_max(end)
      disp('inspect workspace due to non-monotonicity!');
      keyboard;
    end;
    Deltas_max = [Deltas_max, Delta_max];
    model = model.set_mu(model,Mtrain(:,i_max));
    mu_indices = [mu_indices, i_max];
    sim_data = detailed_simulation(model,model_data);
    RB = [RB,model.get_dofs_from_sim_data(sim_data)]; 
    tmp_detailed_data.RB = RB;
    reduced_data =  dictionary_extend_reduced_data(...
        model,reduced_data,tmp_detailed_data);    
  end; % loop over N
%  res.hs_res_norms_max = hs_res_norms_max;  
  detailed_data.RB_info.Mtrain = model.Mtrain;
  detailed_data.RB_info.mu_indices = mu_indices;  
  detailed_data.RB_info.Deltas_max = Deltas_max;
  %%%%%%%%%%%%%%%%%%%%%%%%% End Offline Greedy
end;

detailed_data.RB_info.mus = mus;
detailed_data.RB_info.time = toc;
detailed_data.RB = RB;