basisgen.m

Go to the documentation of this file.

% Script basisgen: demonstration of functionality for basis generation
% and generation of plots for comparing different methods

% B. Haasdonk and S. Habiger 27.7.2009

%step = 1 % single detailed simulation with given data and plot. Run
          % this with varying parameters mu until sure that scheme
          % is stable. Modify dt or the data-functions accordingly,
          % until a nice parameter-domain with uniformly stable
          % detailed scheme is obtained.
%step = 2 % generate dummy reduced basis from single trajectory.
          % simple reduced simulation can also be
          % performed. All results should be visually identical
step = 3 % generate more intelligent reduced basis
%step = 4 % generate and test reduced basis for paper experiments 
%step = 5 % generate plots of adaptive parameter grids
          % generate plots of time/N, max_err/min_err, etc. 

switch step
 case 1 % initialize model and single detailed simulation

  model = init_model('convdiff_model');
  model_data = gen_model_data(model);
  params.mu = [1,0];
  sim_data = detailed_simulation(model,model_data,params);
  params.title = 'detailed simulation';
  params.axis_equal = 1;
  plot_sim_data(model,model_data,sim_data,params);
 
 case 2 % simple RB generation by orthonormalizing trajectory
  
  model = init_model('convdiff_model');
  model_data = gen_model_data(model);
  mu_test   = cellfun(@mean, model.mu_ranges);
  %  mu_test2  = cellfun(@min, para 1. case 3: ms.mu_ranges);
  params.mu = mu_test;
  
  % not: gen_detailed_data is not called here as the following lines do
  % the job:
  sim_data  = detailed_simulation(model, model_data, params);
  UON       = model.orthonormalize(model, model_data, sim_data.U);
  detailed_data.grid = model_data.grid;
  detailed_data.W = model_data.W;
  detailed_data.RB = UON;
  
  reduced_data = gen_reduced_data(model, detailed_data, params);
%  params.N = size(detailed_data.RB,2);
%  reduced_data = extract_reduced_data_subset(reduced_data, params);
  params.mu = mu_test;
  rb_sim_data = rb_simulation(model, reduced_data, params);
%  keyboard;
  rb_sim_data = rb_reconstruction(model, detailed_data, rb_sim_data);

  params.axis_equal = 1;
  params.title = 'reduced simulation result';
  plot_sim_data(model, model_data, rb_sim_data, params);
  params.title = 'detailed simulation result';
  plot_sim_data(model, model_data, sim_data, params);
  RB = detailed_data.RB;
  save(fullfile(rbmatlabresult,'tmp_detailed_data.mat'),'RB','model');
  
 case 3  % test different basis generation methods 
  
  model = init_model('convdiff_model');
  model_data = gen_model_data(model);

%  params.RB_generation_mode  = 'file_load';   
%  params.RB_basis_filename = fullfile(rbmatlabtemp,'tmp_detailed_data.mat');

  params.RB_generation_mode  = 'uniform_fixed';   
  params.RB_numintervals = [4,4];
  params.mu_ranges = model.mu_ranges;
  params.RB_stop_epsilon = 1e-5;
%  params.RB_stop_max_val_train_ratio = ???
  params.RB_stop_timeout = 3600; 
  params.RB_stop_Nmax = 10;
  params.RB_extension_algorithm = @RB_extension_PCA_fixspace;
  params.RB_error_indicator = 'estimator';
%  params.RB_detailed_val_savepath
%  params.RB_detailed_train_savepath =  
  
  %%%%%%%%%%%%%%%%%%% 
 
  detailed_data = gen_detailed_data(model,model_data,params);

  %%%%%%%%%%%%%%%%%%% 
  
  params.title='Reduced Basis';
  params.axis_equal = 1;
  params.clim_tight = 1;
  plot_element_data_sequence(model,model_data.grid,detailed_data.RB,params);
  
  % test different methods of rb_basis_generation
  % X  'file_load' : simple loading o 1. case 3: f reduced basis from a file 
  %
  % X 'uniform_fixed' : greedy algorithm based on a uniform
  %                           cartesian parameter grid
  %  'random_fixed' :  greedy algorithm based on a uniformly
  %                           distributed random parameter set
  %  'PCA_trajectory' : simple detailed simulation and PCA of the
  %                     trajectory
  %  'PCA_trajectories' : simple detailed simulation and PCA of the
  %                     trajectories with parameter vectors specified in 
  %                     params.RB_mu_list
  %  'uniform_refined' :  greedy algorithm based on a uniformly
  %                       random parameter set with grid refinement
  %  'adaptive_refined' : greedy algorithm based on an adaptively 
  %                       refined parameter grid
  
 case 4 % generate different reduced basis by gen_and_test_basis
  
  % the file basis/basisgen/gen_and_test_basis must be updated to
  % new syntax... this is main work

  % ... correct call of gen_and_test_basis with correct arguments
  
  % for example:
  % fns = {'RB_uniform_refined_p3_i2_r1_notimeout_tested',...
  %'RB_uniform_refined_p3_i3_r1_notimeout_tested',...
  %'RB_adaptive_refined_p3_i2_r1_smax1000_theta5e-2_tested',...
  %'RB_adaptive_refined_p3_i3_r1_smax1000_theta5e-2_notimeout_tested',...
  %'RB_uniform_fixed_p3_64_tested',...
  %'RB_uniform_fixed_p3_125_tested'};
  %%'RB_uniform_fixed_p3_216_notimeou
  
  
 case 5 % plots of adaptive grids and error estimator

  % adapt the below code such that new plots are generated...

  % ...
  
  % the following code was used to generate the figures in the 
  % ADMOS Paper on adaptive parameter grids:
  %  figure;
  %load(['C:\Dokumente und Einstellungen\Bernard\Eigene Dateien\',...
  %     'TempMatlab\basisgen\RB_uniform_fixed_64.mat']);
  %params.num_plot_intervals = [40,40];
  %offline_data = rb_offline_prep(detailed_data,params);
  %params.N = size(detailed_data.RB,2);
  %reduced_data = rb_online_prep(offline_data,params);
  %params.no_lines = 1;
  %plot_error_estimator(reduced_data,params);
  %c = get(gca,'Children');
  %cdata = get(c(end),'Cdata');
  %cdata = log10(cdata);
  %set(c(end),'Cdata',cdata);
  %set(gca,'Clim',[-8,-7]);
  %title('log_{10}(\Delta(\mu))');
  %colorbar;
  %%colorbar;
  %hold on;
  %par.numintervals = params.RB_numintervals;
  %par.range = params.mu_ranges; 1. case 3: 
  %par.color = [0,0,0];
  %pgrid = cubegrid(par);
  %plot(pgrid,par);
  %%
  %figure;
  %load(['C:\Dokumente und Einstellungen\Bernard\Eigene Dateien\',...
  %     'TempMatlab\basisgen\RB_uniform_refined_2_2_r1.mat']);
  %params.num_plot_intervals = [40,40];
  %offline_data = rb_offline_prep(detailed_data,params);
  %params.N = size(detailed_data.RB,2);
  %reduced_data = rb_online_prep(offline_data,params);
  %params.no_lines = 1;
  %plot_error_estimator(reduced_data,params);
  %c = get(gca,'Children');
  %cdata = get(c(end),'Cdata');
  %cdata = log10(cdata);
  %set(c(end),'Cdata',cdata);
  %set(gca,'Clim',[-8,-7]);
  %title('log_{10}(\Delta(\mu))');
  %colorbar;
  %hold on;
  %params.no_lines = 0;
  %params.color = [0,0,0];
  %plot(detailed_data.RB_info.MMesh_list{end},params);
  %%CLIM : 2.07876e-008 7.30869e-008
  %%
  %figure;
  %load(['C:\Dokumente und Einstellungen\Bernard\Eigene Dateien\',...
  %     'TempMatlab\basisgen\RB_adaptive_refined_2_2_r1_smax1000.mat']);
  %params.num_plot_intervals = [40,40];
  %offline_data = rb_offline_prep(detailed_data,params);
  %params.N = size(detailed_data.RB,2);
  %reduced_data = rb_online_prep(offline_data,params);
  %params.no_lines = 1;
  %plot_error_estimator(reduced_data,params);
  %c = get(gca,'Children');
  %cdata = get(c(end),'Cdata');
  %cdata = log10(cdata);
  %set(c(end),'Cdata',cdata);
  %set(gca,'Clim',[-8,-7]);
  %title('log_{10}(\Delta(\mu))');
  %colorbar;
  %hold on;
  %params.no_lines = 0;
  %params.color = [0,0,0];
  %plot(detailed_data.RB_info.MMesh_list{end},params);
  %%CLIM : 2.07876e-008 7.30869e-008
  
 case 6 % different plots of time/N, max_err/min_err, etc.
  
  % adapt the below code such that nice plots are generated
  
  % select different basis generation methods
 
  % the following code was used to generate the figures in the 
  % ADMOS Paper on adaptive parameter grids:

  %fns = {'RB_uniform_refined_p3_i2_r1_tested',...
  %'RB_uniform_refined_p3_i3_r1_tested',...
  %'RB_adaptive_refined_p3_i2_r1_smax1000_theta5e-2_tested',...
  %'RB_adaptive_refined_p3_i3_r1_smax1000_theta5e-2_tested',...
  %'RB_uniform_fixed_p3_64_tested',...
  %'RB_uniform_fixed_p3_125_tested',...
  %};
  %
  %params.plot_train_times = 1;
  %params.plot_train_estimators = 0;
  %params.plot_test_estimators = 0;
  %params.plot_test_errors = 0;
  %params.plot_linestyles = {'-','-.','-','-.','-','-.'};
  %params.plot_linecolors = {'r','r',[0,0.6,0],[0,0.6,0],'b','b'};
  %params.plot_linewidths = [2,2,2,2,2,2];
  %params.plot_legends_location = 'Northwest';
  %params.plot_legends = {'uniform-refined 2^3',...
  %                      'uniform-refined 3^3',...
  %                   'adaptive-refined 2^3',...
  %                   'adaptive-refined 3^3',...
  %                   'uniform-fixed 4^3',...
  %                   'uniform-fixed 5^3'};
  %
  % plot_basisgen_results(fns,params);   
  
  % uniform-refined approaches jetzt ohne timeout:
  %fns = {'RB_uniform_refined_p3_i2_r1_notimeout_tested',...
  %'RB_uniform_refined_p3_i3_r1_notimeout_tested',...
  %'RB_adaptive_refined_p3_i2_r1_smax1000_theta5e-2_tested',...
  %'RB_adaptive_refined_p3_i3_r1_smax1000_theta5e-2_notimeout_tested',...
  %'RB_uniform_fixed_p3_64_tested',...
  %'RB_uniform_fixed_p3_125_tested'};
  %'RB_uniform_fixed_p3_216_notimeout_tested',...
  
  %params.plot_train_times = 0;
  %params.plot_train_estimators = 1;
  %params.plot_test_estimators = 0;
  %params.plot_test_errors = 0;
  %params.plot_legends_location = 'NorthEast';
  %plot_basisgen_results(fns,params);
  
  % => Manuelles Speichern der Bilder als 
  %    computation_times.eps und
  %    train_errors.eps

  
  %%%%%% run basisgen_main with the following settings in the header:
  %%
  %%test_RB_methods = {...
  %%    'RB_uniform_refined_2_2_r1',...
  %%    'RB_uniform_refined_2_2_r5',...
  %%    'RB_uniform_fixed_64',...
  %%    'RB_uniform_fixed_256',...
  %%    'RB_uniform_refined_2_2_r1',...
  %%    'RB_uniform_refined_2_2_r5',...
  %%    'RB_adaptive_refined_2_2_r1_smax1000_theta5e-2',...
  %%    'RB_adaptive_refined_2_2_r5_smax1000_theta5e-2',...
  %%    'RB_adaptive_refined_2_2_r1_smax1000_theta1e-1',...
  %%    'RB_adaptive_refined_2_2_r5_smax1000_theta1e-1',...
  %%               };
  %% test_RB_indicators = {'estimator'}; % only estimators
  %% compute_RB_methods = test_RB_methods;
  %
  %%%%%%% then perform the following on the generated files for getting 
  %%%%%%% the figures
  
  %%fns = {...
  %%    'RB_uniform_refined_2_2_r1_tested',...
  %%    'RB_uniform_fixed_256_tested',...
  %%    'RB_adaptive_refined_2_2_r1_smax1000_theta5e-2_tested'};
  
  %% uniform-refined approaches jetzt ohne timeout:
  %fns = {'RB_uniform_refined_p3_i2_r1_notimeout_tested',...
  %'RB_uniform_refined_p3_i3_r1_notimeout_tested',...
  %'RB_adaptive_refined_p3_i2_r1_smax1000_theta5e-2_tested',...
  %'RB_adaptive_refined_p3_i3_r1_smax1000_theta5e-2_notimeout_tested',...
  %'RB_uniform_fixed_p3_64_tested',...
  %'RB_uniform_fixed_p3_125_tested'};
  %%'RB_uniform_fixed_p3_216_notimeout_tested',...
  
  %params.plot_train_times = 0;
  %params.plot_train_estimators = 0;
  %params.plot_test_estimators = 1;
  %params.plot_test_errors = 1;
  %params.plot_linestyles = {'-','-.','-','-.','-','-.'};
  %params.plot_linecolors = {'r','r',[0,0.6,0],[0,0.6,0],'b','b'};
  %params.plot_linewidths = [2,2,2,2,2,2];
  %params.plot_legends_location = 'North';
  %params.plot_legends = {'uniform-refined 2^3',...
  %                      'uniform-refined 3^3',...
  %                   'adaptive-refined 2^3',...
  %                   'adaptive-refined 3^3',...
  %                   'uniform-fixed 4^3',...
  %                   'uniform-fixed 5^3'};
  %
  %plot_basisgen_results(fns,params);
  %close(gcf-1);
  %close(gcf);
  %
  %params.plot_legends_location = 'NorthWest';
  %plot_basisgen_results(fns,params);
  %close(gcf-2);
  %close(gcf-1);
  %
  %% => Manuelles Speichern der Bilder als 
  %%    test_errors.eps und
  %%    test_error_ratio.eps
  
  
  %  fns = {'RB_uniform_refined_p3_i2_r1_notimeout_tested',...
  %'RB_uniform_refined_p3_i3_r1_notimeout_tested',...
  %'RB_adaptive_refined_p3_i2_r1_smax1000_theta5e-2_tested',...
  %'RB_adaptive_refined_p3_i3_r1_smax1000_theta5e-2_notimeout_tested',...
  %'RB_uniform_fixed_p3_64_tested',...
  %'RB_uniform_fixed_p3_125_tested'};
  %%'RB_uniform_fixed_p3_216_notimeout_tested',...
  %
  %params.plot_train_times = 0;
  %params.plot_train_estimators = 0;
  %params.plot_test_estimators = 1;
  %params.plot_test_errors = 0;
  %params.plot_max_test_vs_time = 1;
  %params.plot_linestyles = {'-','-.','-','-.','-','-.'};
  %params.plot_linecolors = {'r','r',[0,0.6,0],[0,0.6,0],'b','b'};
  %params.plot_linewidths = [2,2,2,2,2,2];
  %params.plot_legends_location = 'NorthEast';
  %params.plot_legends = {'uniform-refined 2^3',...
  %                      'uniform-refined 3^3',...
  %                   'adaptive-refined 2^3',...
  %                   'adaptive-refined 3^3',...
  %                   'uniform-fixed 4^3',...
  %                   'uniform-fixed 5^3'};
  %
  %plot_basisgen_results(fns,params);
  %close(gcf-3);
  %close(gcf-2);
  %close(gcf-1);
  %set(gca,'Xlim',[0,3600]);
  %set(gca,'Ylim',[1e-7,2e-4]);
  
 otherwise
  error('step number unknown')
end;



%| \docupdate