Fixes and organization

- Removed unnecessary functions/scripts.
- Bugs and issues fixed:
i) During classification any NaN values will be equal to zero. NaN
values were generated during the internal normalization procedure when
both min and max values of a feature were 0.

ii) Fixed a hard coded variable in features computation.

iii) Fixed the Labelling Quality procedure which was causing an error
upon initiation.

iv) Fixed an issue with the agreement matrix where the numbers were
generated into wrong positions.

Updated code for the Demo procedure.

Author: avouros

clustering/cross_validation/new/cross_validation_classifier.m

@@ -2,5 +2,5 @@ VARIABLE,VALUE,COMMENT
 FontName,Arial,figures and export
 FontSize,25,figures and export
 LineWidth,1,figures and export
-Export,.pdf,export
-ExportStyle,High Quality,export
+Export,.jpg,export
+ExportStyle,Low Quality,export

configs/figure/configs.txt

@@ -2,20 +2,31 @@ function demo(set,user_path,varargin)
 %DEMO executes the processes: segmentation, labelling, classification
 %and produces the results
 
+%How to run standalone (everything, no UI, no display):
+% demo(1,'C:\','WAITBAR',0,'DISPLAY',0,'SEGMENTATION',1,...
+%    'CLASSIFICATION',1,'MCLASSIFICATION',1,'LABELLING_QUALITY',1,'STATISTICS',1,'PROBABILITIES',1);
+
+
+    if isempty(varargin)
+        initialization;
+    else
+        if ~isequal(varargin{1},'NO_INIT')
+            initialization;
+        end
+    end
+
     % Functionalities
     SEGMENTATION = 1;
     CLASSIFICATION = 1;
     MCLASSIFICATION = 1;
-    % Clusters
-    UCLUSTERS = 1;
     % Results
     LABELLING_QUALITY = 0;
     STATISTICS = 1;
     PROBABILITIES = 1;
     % Extra
     WAITBAR = 0;
     DISPLAY = 0;
-    
+
     for i = 1:length(varargin)
         if isequal(varargin{i},'WAITBAR')
             WAITBAR = varargin{i+1};
@@ -39,16 +50,8 @@ function demo(set,user_path,varargin)
     %Set Options
     if set == 1
         seg_overlap = [0.7,0.7,0.9,0.7];
-        seg_length = [300,250,250,200];
+        seg_length = [300,250,250,200];      
         groups = [1,2];
-        if UCLUSTERS
-            num_clusters = {300,0.7,[10:49,53], 250,0.7,[10:68,70,74,75,82,83,84], 250,0.9,10:100, 200,0.7,[10:54,56:60]}; %41,65,91,50 classifiers
-        else
-            num_clusters = 10:58; %49 classifiers
-        end
-        sample = [41,65,91,50];
-        iterations = 1;
-        %Mclusters = num_clusters;
     elseif set == 2
         return
     end
@@ -108,21 +111,23 @@ function demo(set,user_path,varargin)
         delete(h);
     end
 
+    %% Load the settings
+    try
+        load(fullfile(project_path,'settings','new_properties.mat'));
+        load(fullfile(project_path,'settings','animal_groups.mat'));
+        load(fullfile(project_path,'settings','my_trajectories.mat'));   
+        full_trajectory_features(project_path,'WAITBAR',WAITBAR,'DISPLAY',DISPLAY); % Trajectories
+        load(fullfile(project_path,'settings','my_trajectories_features.mat'));
+    catch
+        errordlg('Cannot load project settings','Error');
+        return
+    end        
+    
     %% Segmentation
     if SEGMENTATION
-        try
-            load(fullfile(project_path,'settings','new_properties.mat'));
-            load(fullfile(project_path,'settings','animal_groups.mat'));
-            load(fullfile(project_path,'settings','my_trajectories.mat'));
-        catch
-            errordlg('Cannot load project settings','Error');
-            return
-        end
-        full_trajectory_features(project_path); % Trajectories
-        load(fullfile(project_path,'settings','my_trajectories_features.mat'));
         for i = 1:length(seg_overlap)
             seg_properties = [seg_length(i),seg_overlap(i)];
-            segmentation_configs = config_segments(new_properties, seg_properties, trajectory_groups, my_trajectories, my_trajectories_features, '');
+            segmentation_configs = config_segments(new_properties, seg_properties, trajectory_groups, my_trajectories, my_trajectories_features, '', 'WAITBAR',WAITBAR,'DISPLAY',DISPLAY);
             save_segmentation(segmentation_configs, project_path);
         end    
     end
@@ -145,7 +150,7 @@ function demo(set,user_path,varargin)
             end
         end
     end   
-    
+
     %% Animal Trajectories Map
     [~, animals_trajectories_map, animals_ids] = trajectories_map(my_trajectories,my_trajectories_features,groups,'Friedman',set);
 
@@ -166,15 +171,13 @@ function demo(set,user_path,varargin)
                         results_strategies_distributions_manual_full(project_path,tmp1,tmp2,animals_trajectories_map,1);
                         break;
                     end
-                    if iscell(num_clusters)
-                        for k = 1:3:length(num_clusters)
-                            if num_clusters{k} == str2num(seg_length) && num_clusters{k+1} == str2num(seg_overlap)/10
-                                error = execute_classification(project_path,sfiles(j).name,lfiles(i).name,num_clusters{k+2},varargin);
-                            end
-                        end
-                    else
-                        error = execute_classification(project_path,sfiles(j).name,lfiles(i).name,num_clusters,varargin);    
-                    end
+                    %read the cv file
+                    cv_file = fullfile(datapath,'cv',strcat('cv_',seg_length,'_',seg_overlap,'.csv'));
+                    data = read_cv_file(cv_file);
+                    data = data(:,[1,2,5]);
+                    cvErr = find(data(:,3) < 25);
+                    num_clusters = data(cvErr,1);    
+                    error = execute_classification(project_path,sfiles(j).name,lfiles(i).name,num_clusters,varargin);
                 end
             end        
         end
@@ -186,7 +189,7 @@ function demo(set,user_path,varargin)
         for i = 3:length(classifs)
             %execute_Mclassification(project_path, {classifs(i).name}, sample, iterations, 0, 'CLUSTERS', Mclusters)
             tmp = dir(fullfile(project_path,'classification',classifs(i).name,'*.mat'));
-            execute_Mclassification(project_path, {classifs(i).name}, length(tmp), iterations, 0, varargin{:});
+            execute_Mclassification(project_path, {classifs(i).name}, length(tmp), 1, 0, varargin{:});
         end
     end   
 
@@ -205,7 +208,7 @@ function demo(set,user_path,varargin)
         errordlg('Error: metrics generation','Error');
     end
 
-    % STRATEGIES - TRANSITIONS - PROBABILITIES - STATISTICS
+    % STRATEGIES - TRANSITIONS - PROBABILITIES - STATISTICS (Ensemble)
     if PROBABILITIES
         b_pressed = {'Strategies','Transitions','Probabilities'};
     else
@@ -259,6 +262,60 @@ function demo(set,user_path,varargin)
         end
     end
 
+    % STRATEGIES - TRANSITIONS - PROBABILITIES - STATISTICS (Classifiers)
+    if PROBABILITIES
+        b_pressed = {'Strategies','Transitions','Probabilities'};
+    else
+        b_pressed = {'Strategies','Transitions'};
+    end
+    segs = dir(fullfile(project_path,'segmentation','*.mat'));
+    cclasses = dir(fullfile(project_path,'classification'));
+    for i = 1:length(segs)
+        seg = fullfile(project_path,'segmentation',segs(i).name);
+        [~,~,sl,so] = split_segmentation_name(seg);
+        if isequal(sl,'0') || isequal(so,'0')
+            continue;
+        end        
+        for j = 3:length(cclasses)
+            mclass = fullfile(project_path,'classification',cclasses(j).name);
+            [~,~,~,seg_length,seg_overlap,~,~,~] = split_mclassification_name(mclass);
+            if isequal(sl,seg_length) && isequal(so,seg_overlap)
+                load(fullfile(project_path,'segmentation',segs(i).name));
+                % Check the classification
+                [error,name,classifications] = check_classification(project_path,segmentation_configs,cclasses(j).name, 'WAITBAR',WAITBAR, 'DISPLAY', DISPLAY);
+                if error
+                    errordlg('Classification check failed','Error');
+                    return
+                end    
+                % Generate the results
+                for b = 1:length(b_pressed)
+                    error = generate_results(project_path, name, my_trajectories, segmentation_configs, classifications, animals_trajectories_map, animals_ids, b_pressed{b}, groups, 'WAITBAR',WAITBAR, 'DISPLAY', DISPLAY);
+                    if error
+                        errordlg('Cannot create results for strategies, transitions and probabilities','Error');
+                        return
+                    end
+                    error = generate_results(project_path, name, my_trajectories, segmentation_configs, classifications, animals_trajectories_map, animals_ids, b_pressed{b}, groups, 'DISTRIBUTION',2, 'EXTRA_NAME','_nosmooth', 'WAITBAR',WAITBAR, 'DISPLAY', DISPLAY);
+                    if error
+                        errordlg('Cannot create results for strategies, transitions and probabilities (smooth)','Error');
+                        return
+                    end                    
+                end     
+                % Statistics
+                if STATISTICS
+                    [error,~,~] = class_statistics(project_path, cclasses(j).name, 'WAITBAR', WAITBAR);
+                    if error
+                        errordlg('Error: statistics generation','Error');
+                    end  
+                    [error,~,~] = class_statistics(project_path, cclasses(j).name, 'SEGMENTATION', segmentation_configs, 'WAITBAR', WAITBAR, 'DISPLAY', DISPLAY);
+                    if error
+                        errordlg('Error: statistics generation (smooth)','Error');
+                    end
+                end
+                break;
+            end
+        end
+    end
+    
     %% Labelling Quality
     if LABELLING_QUALITY
         for i = 1:length(seg_overlap)

results/demo/demo.m renamed to demo.m

@@ -55,7 +55,15 @@ function ok_button_Callback(hObject, eventdata, handles)
     PROBABILITIES = get(handles.opt_probabilities,'Value');
     WAITBAR = get(handles.opt_wait,'Value');
     DISPLAY = get(handles.opt_display,'Value');
-    demo(set,user_path,'LABELLING_QUALITY',LABELLING_QUALITY,'STATISTICS',STATISTICS,...
+    handles.set_menu.Enable = 'off';
+    handles.opt_wait.Enable = 'off';
+    handles.opt_display.Enable = 'off';
+    handles.opt_quality.Enable = 'off';
+    handles.opt_statistics.Enable = 'off';
+    handles.opt_probabilities.Enable = 'off';
+    handles.cancel_button.Enable = 'off';
+    handles.ok_button.Enable = 'off';
+    demo(set,user_path,'NO_INIT','LABELLING_QUALITY',LABELLING_QUALITY,'STATISTICS',STATISTICS,...
         'PROBABILITIES',PROBABILITIES,'WAITBAR',WAITBAR,'DISPLAY',DISPLAY);
     demo_gui_CloseRequestFcn(hObject, eventdata, handles);
 

demo_gui.m

@@ -12,6 +12,8 @@
 
     d = size(x);
     flag = 1; % capture any error
+	tmp = find(isnan(x));
+    x(tmp) = 0
 
      % create attributes Cell Array
     attr = {};
@@ -30,12 +32,12 @@
     mpck.setTotalTrainWithLabels(data);    
 
     % in case of error continue;
-    try
+    %try
         mpck.buildClusterer(c, data, data, k, data.numInstances())   
-    catch
+    %catch
         flag = 0;
         %disp('error');
-    end    
+    %end    
 
     distr = mpck.getClusterAssignments();
     distr = distr';