pls options added

fluents/lib/blmfuncs.py

@@ -1,9 +1,10 @@
 """This module contains bilinear models(Functions)
 """
-
+import os
 import pygtk
 import gtk
 import gtk.glade
+import fluents
 from fluents.workflow import Function, OptionsDialog, Options
 from fluents.dataset import Dataset
 from fluents import plots, dataset, workflow, logger
@@ -75,6 +76,9 @@ class PCA(Model):
         
         jk_segments = pca_jkP(self.model['E0'], aopt, n_sets)
         Pcal = self.model['P'][:,:aopt]
+        # add the scale to P
+        tnorm = scipy.apply_along_axis(norm, 0, self.model['T'])
+        Pcal = Pcal*tnorm
         tsq = hotelling(jk_segments, Pcal, p_center,
                         cov_center, alpha, crot, strict)
         self.model['p_tsq'] = tsq
@@ -95,7 +99,7 @@ class PCA(Model):
         #####
         if do_lev_s:
             # sample leverages
-            tnorm = scipy.apply_along_axis(norm, 0, dat['T']) # norm of T-columns
+            tnorm = scipy.apply_along_axis(norm, 0, dat['T']) # norm of Ts
             s_lev = leverage(amax, tnorm)
             dat['s_lev'] = s_lev
         if do_lev_v:
@@ -191,13 +195,22 @@ class PLS(Model):
         Model.__init__(self, id, name)
         self._options = PlsOptions()
         
-    def pre_validation(self, amax, n_sets, val_engine):
+    def validation(self, amax, n_sets, cv_val_method):
         """Returns rmsec,rmsep for model.
         """
-        rmsep, aopt = val_engine(self.model['E0'], self.model['F0'],
+        m, n = self.model['E0'].shape
-                           amax, n_sets)
+        if m>n:
-        self.model['rmsep'] = rmsep.mean(0)
+            val_engine = w_pls_cv_val
-        self.model['aopt'] = aopt
+        else:
+            val_engine = pls_val
+        if self._options['calc_cv']==True:
+            rmsep, aopt = val_engine(self.model['E0'], self.model['F0'],
+                                     amax, n_sets)
+            self.model['rmsep'] = rmsep[:,:-1]
+            self.model['aopt'] = aopt
+        else:
+            self.model['rmsep'] = None
+            self.model['aopt'] = self._options['aopt']
         
     def confidence(self, aopt, n_sets, alpha, p_center,
                    crot, strict, cov_center ):
@@ -205,21 +218,35 @@ class PLS(Model):
         Supported parameters: W
         """
         aopt = self.model['aopt']
-        jk_segments = pls_jkW(self.model['E0'], self.model['F0'],
+        if self._options['calc_conf']:
-                              aopt, n_sets)
+            jk_segments = pls_jkW(self.model['E0'], self.model['F0'],
-        Wcal = self.model['W'][:,:aopt]
+                                  aopt, n_sets)
-        tsq = hotelling(jk_segments, Wcal, p_center,
+            Wcal = self.model['W'][:,:aopt]
-                              alpha, crot, strict, cov_center)
+            # ensure that Wcal is scaled
-        self.model['w_tsq'] = tsq
+            tnorm = scipy.apply_along_axis(norm, 0, self.model['T'][:,:aopt])
+            Wcal = Wcal*tnorm
+            tsq = hotelling(jk_segments, Wcal, p_center,
+                            alpha, crot, strict, cov_center)
+            self.model['w_tsq'] = tsq
+        else:
+            self.model['w_tsq'] = None
 
     def permutation_confidence(self, a, b, aopt, reg, n_iter, algo,
                                sim_method):
-        """Estimates sign. vars by controlling fdr."""
+        """Estimates cut off on significant vars by controlling fdr."""
-
-        qvals_sorted, qvals = pls_qvals(a, b, aopt=None,
-                                        alpha=.4, n_iter=20, algo='pls',
-                                        sim_method='shuffle', )
 
+        if self._options['calc_qvals']==True:
+            qvals_sorted, qvals = pls_qvals(a, b,
+                                            aopt=None,
+                                            alpha=reg,
+                                            n_iter=n_iter,
+                                            algo='pls',
+                                            sim_method=sim_method)
+            self.model['qval'] = qvals
+            self.model['qval_sorted'] = qvals_sorted
+        else:
+           self.model['qval'] = None
+           self.model['qval_sorted'] = None 
 
     def make_model(self, a, b, amax, scale, mode, engine):
         """Make model on amax components.
@@ -243,7 +270,8 @@ class PLS(Model):
         # y vars
         ids_3 = [dim_name_3, DY.get_identifiers(dim_name_3, sorted=True)]
         # components (hidden)
-        pc_ids = ['_comp', map(str, range(self.model['aopt']))]
+        pc_ids = ['_comp', map(str, range(self._options['amax']))]
+        pc_ids_opt = ['_comp', map(str, range(self.model['aopt']))]
         zero_dim = ['_doe',['0']] # null dim, vector (hidden)
 
         match_ids = {'E':[ids_0, ids_1],
@@ -257,11 +285,11 @@ class PLS(Model):
                      'qval':[ids_1, zero_dim],
                      'qval_sorted':[ids_1, zero_dim],
                      'w_tsq':[ids_1, zero_dim],
-                     'rmsep':[pc_ids, zero_dim],
+                     'rmsep':[ids_3, pc_ids],
                      }
         
         array = self.model[name]     
-        M = Dataset(array,identifiers=match_ids[name],name=name)
+        M = Dataset(array, identifiers=match_ids[name], name=name)
         return M
 
     def get_out_plots(self, options):
@@ -274,11 +302,13 @@ class PLS(Model):
         return out
             
     def run_o(self, a, b):
+        """Run PLS with present options."""
         options = self._options
         self._dataset['X'] = a
         self._dataset['Y'] = b
         self._data['X'] = a.asarray()
         self._data['Y'] = b.asarray()
+
         if options['center']:
             self.model['E0'] = options['center_mth'](self._data['X'])
             self.model['F0'] = options['center_mth'](self._data['Y'])
@@ -286,7 +316,7 @@ class PLS(Model):
             self.model['E0'] = self._data['X']
             self.model['F0'] = self._data['Y']
         
-        self.pre_validation(**options.pre_validation_options())
+        self.validation(**options.validation_options())
         self.make_model(self.model['E0'], self.model['F0'],
                         **options.make_model_options())
         # variance captured
@@ -307,7 +337,7 @@ class PLS(Model):
         return out
 
     def run(self, a, b):
-        """Run Pls with option gui.
+        """Run PLS with option gui.
         """
         dialog = PlsOptionsDialog([a, b], self._options)
         dialog.show_all()
@@ -319,6 +349,8 @@ class PLS(Model):
             dialog.set_output()
             
             #run with current data and options
+            for key, val in self._options.items():
+                print (key, val)
             return self.run_o(a, b)
 
 class Packer:
@@ -342,7 +374,7 @@ class Packer:
 
         if axis == 1:
             self._array = self._array.T
-        u,s,vt = svd(self._array,full_matrices=0)
+        u, s, vt = svd(self._array,full_matrices=0)
         self._inflater = vt.T
         self._packed_data = u*s
         return self._packed_data
@@ -392,10 +424,11 @@ class PcaOptions(Options):
                            ('rmsep', 'RMSEP', False)
                            ]
 
+        # (class, name, sensitive, ticked)
         opt['all_plots'] = [(blmplots.PcaScorePlot, 'Scores', True),
                             (blmplots.PcaLoadingPlot, 'Loadings', True),
                             (blmplots.LineViewXc, 'Line view', True),
-                            (blmplots.PredictionErrorPlot, 'Residual Error', True)
+                            (blmplots.PredictionErrorPlot, 'Residual Error', False)
                             ]
         
         opt['out_data'] = ['T','P', 'p_tsq']
@@ -405,7 +438,7 @@ class PcaOptions(Options):
         
     def make_model_options(self):
         """Options for make_model method."""
-        opt_list = ['scale','mode', 'amax']
+        opt_list = ['scale', 'mode', 'amax']
         return self._copy_from_list(opt_list)   
 
     def confidence_options(self):
@@ -433,24 +466,24 @@ class PlsOptions(Options):
         opt['algo'] = 'pls'
         opt['engine'] = engines.pls
         opt['mode'] = 'normal' # how much info to calculate
-        opt['lod'] = 'compact' # how much info to store
+        opt['amax'] = 10
-        opt['amax'] = 3
+        opt['aopt'] = 10
-        opt['aopt'] = 3
+        opt['auto_aopt'] = False
         opt['center'] = True
         opt['center_mth'] = mat_center
         opt['scale'] = 'scores'
-        opt['calc_conf'] = False
-        opt['n_sets'] = 10
 
+        opt['calc_conf'] = False
+        opt['n_sets'] = 5
         opt['strict'] = True
         opt['p_center'] = 'med'
-        opt['alpha'] = .2
+        opt['alpha'] = .8
         opt['cov_center'] = 'med'
         opt['crot'] = True
 
-        opt['calc_cv'] = True
+        opt['calc_cv'] = False
-        opt['calc_pert'] = False
+        opt['cv_val_method'] = 'random'
-        opt['val_engine'] = w_pls_cv_val
+        opt['cv_val_sets'] = opt['n_sets']
 
         opt['all_data'] = [('T', 'scores', True),
                            ('P', 'loadings', True),
@@ -459,18 +492,22 @@ class PlsOptions(Options):
                            ('rmsep', 'RMSEP', False)
                            ]
 
+        # (class, name, sensitive, ticked)
         opt['all_plots'] = [(blmplots.PlsScorePlot, 'Scores', True),
                             (blmplots.PlsLoadingPlot, 'Loadings', True),
-                            (blmplots.LineViewXc, 'Line view', True)]
+                            (blmplots.LineViewXc, 'Line view', True),
+                            (blmplots.PredictionErrorPlot, 'Residual Error', False),
+                            (blmplots.RMSEPPlot, 'RMSEP', False)
+                            ]
+        
+        opt['out_data'] = ['T','P', 'p_tsq']
+        opt['out_plots'] = [blmplots.PlsScorePlot,blmplots.PlsLoadingPlot,blmplots.LineViewXc]
         
-        opt['out_plots'] = [blmplots.PlsScorePlot,
-                             blmplots.PlsLoadingPlot,
-                            blmplots.LineViewXc]
         opt['out_data'] = None
                             
         opt['pack'] = False
         opt['calc_qvals'] = False
-        opt['q_pert_mth'] = 'shuffle_vars'
+        opt['q_pert_method'] = 'shuffle_rows'
         opt['q_iter'] = 20        
 
         self.update(opt)
@@ -486,19 +523,22 @@ class PlsOptions(Options):
                     'strict', 'crot', 'cov_center'] 
         return self._copy_from_list(opt_list)  
     
-    def pre_validation_options(self):
+    def validation_options(self):
         """Options for pre_validation method."""
-        opt_list = ['amax', 'n_sets', 'val_engine']
+        opt_list = ['amax', 'n_sets', 'cv_val_method']
         return self._copy_from_list(opt_list)
 
+    def permutation_confidence(self):
+        opt_list = ['q_pert_method', 'q_iter']
+        return self._copy_from_list(opt_list)
 
 class PcaOptionsDialog(OptionsDialog):
     """Options dialog for Principal Component Analysis.
     """
     def __init__(self, data, options, input_names=['X']):
         OptionsDialog.__init__(self, data, options, input_names)
-
+        #glade_file = os.path.join(fluents.DATADIR, 'pca_options.glade')
-        glade_file = os.path.join(fluents.DATADIR, 'pca_options.glade')
+        glade_file = os.path.join("/home/flatberg/fluents/fluents/", 'pca_options.glade')
 
         notebook_name = "vbox1"
         page_name = "Options"
@@ -508,19 +548,24 @@ class PcaOptionsDialog(OptionsDialog):
                "on_aopt_value_changed" : self.on_aopt_changed,
                "auto_aopt_toggled" : self.auto_aopt_toggled,
                "center_toggled" : self.center_toggled,
-               "on_scale_changed" : self.on_scale_changed,
+               #"on_scale_changed" : self.on_scale_changed,
                "on_val_none" : self.val_toggled,
                "on_val_cv" : self.cv_toggled,
                "on_val_pert" : self.pert_toggled,
                "on_cv_method_changed" : self.on_cv_method_changed,
                "on_cv_sets_changed" : self.on_cv_sets_changed,
                "on_pert_sets_changed" : self.on_pert_sets_changed,
-               "on_conf_toggled" : self.on_conf_toggled
+               "on_conf_toggled" : self.on_conf_toggled,
+               "on_subset_loc_changed" : self.on_subset_loc_changed,
+               "on_cov_loc_changed" : self.on_cov_loc_changed,
+               "on_alpha_changed" : self.on_alpha_changed,
+               "on_rot_changed" : self.on_rot_changed
                }
         
         self.wTree.signal_autoconnect(dic)
 
         # set/ensure valid default values/ranges
+        #
         amax_sb = self.wTree.get_widget("amax_spinbutton")
         max_comp =  min(data[0].shape) # max num of components
         if self._options['amax']>max_comp:
@@ -534,8 +579,8 @@ class PcaOptionsDialog(OptionsDialog):
         aopt_sb.get_adjustment().set_all(self._options['aopt'], 1, self._options['amax'], 1, 0, 0)
 
         # scale
-        scale_cb = self.wTree.get_widget("scale_combobox")
+        # scale_cb = self.wTree.get_widget("scale_combobox")
-        scale_cb.set_active(0)
+        # scale_cb.set_active(0)
 
         # validation frames
         if self._options['calc_cv']==False:
@@ -550,9 +595,20 @@ class PcaOptionsDialog(OptionsDialog):
 
         # confidence
         if self._options['calc_conf']==True:
-            self.wTree.get_widget("subset_frame").set_sensitive(True)
+            self.wTree.get_widget("subset_expander").set_sensitive(True)
         else:
-            self.wTree.get_widget("subset_frame").set_sensitive(False)
+            self.wTree.get_widget("subset_expander").set_sensitive(False)
+
+        cb = self.wTree.get_widget("subset_loc")
+        _m = {'med': 0, 'mean': 1, 'full_model': 2}
+        cb.set_active(_m.get(self._options['p_center']))
+
+        cb = self.wTree.get_widget("cov_loc")
+        _m = {'med': 0, 'mean': 1}
+        cb.set_active(_m.get(self._options['cov_center']))
+
+        hs = self.wTree.get_widget("alpha_scale")
+        hs.set_value(self._options['alpha'])
         
                     
     def on_amax_changed(self, sb):
@@ -585,14 +641,14 @@ class PcaOptionsDialog(OptionsDialog):
             logger.log("debug", "centering set to False")
             self._options['center'] = False
 
-    def on_scale_changed(self, cb):
+    #def on_scale_changed(self, cb):
-        scale = cb.get_active_text()
+    #    scale = cb.get_active_text()
-        if scale=='Scores':
+    #    if scale=='Scores':
-            self._options['scale'] = 'scores'
+    #        self._options['scale'] = 'scores'
-        elif scale=='Loadings':
+    #    elif scale=='Loadings':
-            self._options['scale'] = 'loads'
+    #        self._options['scale'] = 'loads'
-        else:
+    #    else:
-            raise IOError
+    #        raise IOError
 
     def val_toggled(self, tb):
         """Callback for validation: None. """
@@ -657,17 +713,242 @@ class PcaOptionsDialog(OptionsDialog):
     def on_conf_toggled(self, tb):
         if tb.get_active():
             self._options['calc_conf'] = False
-            self.wTree.get_widget("subset_frame").set_sensitive(False)
+            self.wTree.get_widget("subset_expander").set_sensitive(False)
         else:
             self._options['calc_conf'] = True
-            self.wTree.get_widget("subset_frame").set_sensitive(True)
+            self.wTree.get_widget("subset_expander").set_sensitive(True)
 
+    def on_subset_loc_changed(self, cb):
+        method = cb.get_active_text()
+        if method=='Full model':
+            self._options['p_center'] = 'full_model'
+        elif method=='Median':
+            self._options['p_center'] = 'med'
+        elif method=='Mean':
+            self._options['p_center'] = 'mean'
 
+    def on_cov_loc_changed(self, cb):
+        method = cb.get_active_text()
+        if method=='Median':
+            self._options['cov_center'] = 'med'
+        elif method=='Mean':
+            self._options['cov_center'] = 'mean'
+
+    def on_alpha_changed(self, hs):
+        self._options['alpha'] = hs.get_value()
+
+    def on_rot_changed(self, rg):
+        proc, strict = rg
+        if proc.get_active():
+            self._options['crot'] = True
+        else:
+            self._options['crot'] = True
+            self._options['strict'] = True 
 
 
 class PlsOptionsDialog(OptionsDialog):
-    """Options dialog for Partial Least Squares Regression.
+    """Options dialog for Partial Least squares regression.
     """
-
     def __init__(self, data, options, input_names=['X', 'Y']):
         OptionsDialog.__init__(self, data, options, input_names)
+        #glade_file = os.path.join(fluents.DATADIR, 'pca_options.glade')
+        glade_file = os.path.join("/home/flatberg/fluents/fluents/", 'pls_options.glade')
+
+        notebook_name = "vbox1"
+        page_name = "Options"
+        self.add_page_from_glade(glade_file, notebook_name, page_name)
+        # connect signals to handlers
+        dic = {"on_amax_value_changed" : self.on_amax_changed,
+               "on_aopt_value_changed" : self.on_aopt_changed,
+               "auto_aopt_toggled" : self.auto_aopt_toggled,
+               "center_toggled" : self.center_toggled,
+               #"on_scale_changed" : self.on_scale_changed,
+               "on_val_none" : self.val_toggled,
+               "on_val_cv" : self.cv_toggled,
+               "on_cv_method_changed" : self.on_cv_method_changed,
+               "on_cv_sets_changed" : self.on_cv_sets_changed,
+               "on_conf_toggled" : self.conf_toggled,
+               "on_subset_loc_changed" : self.on_subset_loc_changed,
+               "on_cov_loc_changed" : self.on_cov_loc_changed,
+               "on_alpha_changed" : self.on_alpha_changed,
+               "on_rot_changed" : self.on_rot_changed,
+               "on__toggled" : self.conf_toggled,
+               "on_qval_changed" : self.on_qval_changed,
+               "on_iter_changed" : self.on_iter_changed
+               }
+        
+        self.wTree.signal_autoconnect(dic)
+
+        # set/ensure valid default values/ranges
+        #
+        amax_sb = self.wTree.get_widget("amax_spinbutton")
+        max_comp =  min(data[0].shape) # max num of components
+        if self._options['amax']>max_comp:
+            logger.log('debug', 'amax default too large ... adjusting')
+            self._options['amax'] = max_comp
+        amax_sb.get_adjustment().set_all(self._options['amax'], 1, max_comp, 1, 0, 0)
+        # aopt spin button
+        aopt_sb = self.wTree.get_widget("aopt_spinbutton")
+        if self._options['aopt']>self._options['amax']:
+            self._options['aopt'] = self._options['amax'] + 1 - 1
+        aopt_sb.get_adjustment().set_all(self._options['aopt'], 1, self._options['amax'], 1, 0, 0)
+
+        # scale
+        # scale_cb = self.wTree.get_widget("scale_combobox")
+        # scale_cb.set_active(0)
+
+        # validation frames
+        if self._options['calc_cv']==False:
+            cv_frame = self.wTree.get_widget("cv_frame")
+            cv_frame.set_sensitive(False)
+
+        cv = self.wTree.get_widget("cv_method").set_active(0)
+
+        # confidence
+        if self._options['calc_conf']==True:
+            self.wTree.get_widget("subset_expander").set_sensitive(True)
+        else:
+            self.wTree.get_widget("subset_expander").set_sensitive(False)
+
+        cb = self.wTree.get_widget("subset_loc")
+        _m = {'med': 0, 'mean': 1, 'full_model': 2}
+        cb.set_active(_m.get(self._options['p_center']))
+
+        cb = self.wTree.get_widget("cov_loc")
+        _m = {'med': 0, 'mean': 1}
+        cb.set_active(_m.get(self._options['cov_center']))
+
+        hs = self.wTree.get_widget("alpha_scale")
+        hs.set_value(self._options['alpha'])
+
+        tb = self.wTree.get_widget("qvals")
+        tb.set_sensitive(True)
+        
+                    
+    def on_amax_changed(self, sb):
+        logger.log("debug", "amax changed: new value: %s" %sb.get_value_as_int())
+        amax = sb.get_value_as_int()
+        # update aopt if needed
+        if amax<self._options['aopt']:
+            self._options['aopt'] = amax
+        aopt_sb = self.wTree.get_widget("aopt_spinbutton")
+        aopt_sb.get_adjustment().set_all(self._options['aopt'], 1, amax, 1, 0, 0)
+        self._options['amax'] = sb.get_value_as_int()
+        
+    def on_aopt_changed(self, sb):
+        aopt = sb.get_value_as_int()
+        self._options['aopt'] = aopt
+        
+    def auto_aopt_toggled(self, tb):
+        aopt_sb = self.wTree.get_widget("aopt_spinbutton")
+        if tb.get_active():
+            self._options['auto_aopt'] = True
+            aopt_sb.set_sensitive(False)
+        else:
+            self._options['auto_aopt'] = False
+            aopt_sb.set_sensitive(True)
+
+    def center_toggled(self, tb):
+        if tb.get_active():
+            self._options['center'] = True
+        else:
+            logger.log("debug", "centering set to False")
+            self._options['center'] = False
+
+    #def on_scale_changed(self, cb):
+    #    scale = cb.get_active_text()
+    #    if scale=='Scores':
+    #        self._options['scale'] = 'scores'
+    #    elif scale=='Loadings':
+    #        self._options['scale'] = 'loads'
+    #    else:
+    #        raise IOError
+
+    def val_toggled(self, tb):
+        """Callback for validation: None. """
+        cv_frame = self.wTree.get_widget("cv_frame")
+        cv_tb = self.wTree.get_widget("cv_toggle")
+        if tb.get_active():
+            self._options['calc_cv'] = False
+            cv_frame.set_sensitive(False)
+            cv_tb.set_sensitive(False)
+        else:
+            cv_tb.set_sensitive(True)
+            if cv_tb.get_active():
+                cv_frame.set_sensitive(True)
+                self._options['calc_cv'] = True
+
+    def cv_toggled(self, tb):
+        cv_frame = self.wTree.get_widget("cv_frame")
+        val_tb = self.wTree.get_widget("val_none_toggle")
+        if tb.get_active():
+            cv_frame.set_sensitive(True)
+            self._options['calc_cv'] = True
+        else:
+            cv_frame.set_sensitive(False)
+            self._options['calc_cv'] = False
+
+    def on_cv_method_changed(self, cb):
+        method = cb.get_active_text()
+        if method == 'Random':
+            self._options['cv_val_method'] = 'random'
+
+    def on_cv_sets_changed(self, sb):
+        val = sb.get_value_as_int()
+        self._options['cv_val_sets'] = val
+
+    def conf_toggled(self, tb):
+        if tb.get_active():
+            self._options['calc_conf'] = False
+            self.wTree.get_widget("subset_expander").set_sensitive(False)
+        else:
+            self._options['calc_conf'] = True
+            self.wTree.get_widget("subset_expander").set_sensitive(True)
+
+    def on_subset_loc_changed(self, cb):
+        method = cb.get_active_text()
+        if method=='Full model':
+            self._options['p_center'] = 'full_model'
+        elif method=='Median':
+            self._options['p_center'] = 'med'
+        elif method=='Mean':
+            self._options['p_center'] = 'mean'
+
+    def on_cov_loc_changed(self, cb):
+        method = cb.get_active_text()
+        if method=='Median':
+            self._options['cov_center'] = 'med'
+        elif method=='Mean':
+            self._options['cov_center'] = 'mean'
+
+    def on_alpha_changed(self, hs):
+        self._options['alpha'] = hs.get_value()
+
+    def on_rot_changed(self, rg):
+        proc, strict = rg
+        if proc.get_active():
+            self._options['crot'] = True
+        else:
+            self._options['crot'] = True
+            self._options['strict'] = True 
+
+    def qval_toggled(self, tb):
+        if tb.get_active():
+            self._options['calc_qval'] = False
+            print "Setting not sens"
+            self.wTree.get_widget("qval_method").set_sensitive(False)
+            self.wTree.get_widget("q_iter").set_sensitive(False)
+        else:
+            self._options['calc_qval'] = True
+            self.wTree.get_widget("qval_method").set_sensitive(True)
+            self.wTree.get_widget("q_iter").set_sensitive(True)
+
+    def on_iter_changed(self, sb):
+        self._options['q_iter'] = sb.get_value()
+
+    def on_qval_changed(self, cb):
+        q_method = cb.get_active_text()
+        if method=='Shuffle rows':
+            self._options['q_pert_method'] = 'shuffle'
+        
+