Changeset 5322

Timestamp:

Aug 24, 2022 3:39:37 PM (16 months ago)

Author:

vondreele

Message:

improvements to scipy cluster analysis & start on Scikit-learn version

Location:

trunk

Files:

• GSASIIdataGUI.py (modified) (2 diffs)
• GSASIIplot.py (modified) (6 diffs)
• GSASIIseqGUI.py (modified) (4 diffs)

trunk/GSASIIdataGUI.py

@@ -5697 +5697 @@
         '''
         
+        try:
+            SKLearn = False
+            import sklearn.cluster
+            SKLearn = True
+        except:
+            res = GSASIIpath.condaInstall('scikit-learn')
+            if res:
+                msg = 'Installation of the sklearn package failed with error:\n' + str(res)
+                G2G.G2MessageBox(self,msg,'Install sklearn Error')
         Id = GetGPXtreeItemId(self,self.root,'Cluster Analysis')
         if not Id:
+                
             Id = self.GPXtree.AppendItem(self.root,text='Cluster Analysis')
             ClustDict = {'Files':[],'Method':'correlation','Limits':[0.,100.],'DataMatrix':[],'plots':'All',
@@ -5704 +5714 @@
             self.GPXtree.SetItemPyData(Id,ClustDict)
         else:
+            ClustDict = self.GPXtree.GetItemPyData(Id)
             print('Cluster Analysis exists - nothing done')                   
-
+        ClustDict['SKLearn'] = SKLearn
         self.GPXtree.SelectItem(Id)
     

trunk/GSASIIplot.py

@@ -11548 +11548 @@
     Font = Page.GetFont()
     cb = wx.ComboBox(G2frame.G2plotNB.status,style=wx.CB_DROPDOWN|wx.CB_READONLY,choices=choice,
-                         size=(G2frame.G2plotNB.status.firstLen,-1))
+        size=(G2frame.G2plotNB.status.firstLen,-1))
     cb.Bind(wx.EVT_COMBOBOX, OnKeyBox)
     text = [str(Layers['Layers'][seq]['Name']) for seq in laySeq]
@@ -11568 +11568 @@
 #### Plot Cluster Analysis ####################################################
 
-# def PlotDendogram(G2frame,CLuDict,CLuZ,newPlot=True):
-    
-#     import scipy.cluster.hierarchy as SCH
-#     global Plot
-#     def OnMotion(event):
-#         xpos = event.xdata
-#         if xpos:                                        #avoid out of frame mouse position
-#             ypos = event.ydata
-#             SetCursor(Page)
-#             try:
-#                 G2frame.G2plotNB.status.SetStatusText('X =%9.3f %s =%9.3g'%(xpos,Title,ypos),1)                   
-#             except TypeError:
-#                 G2frame.G2plotNB.status.SetStatusText('Select '+Title+' pattern first',1)
-#     xylim = []
-#     Title = 'Cluster dendogram'
-#     new,plotNum,Page,Plot,lim = G2frame.G2plotNB.FindPlotTab(Title,'mpl')
-#     if not new:
-#         if not newPlot:
-#             xylim = copy.copy(lim)
-#     else:
-#         newPlot = True
-#         Page.canvas.mpl_connect('motion_notify_event', OnMotion)
-    
-#     Page.Choice = None
-#     G2frame.G2plotNB.status.DestroyChildren() #get rid of special stuff on status bar
-#     Plot.set_title('%s %s'%(CLuDict['LinkMethod'],Title))
-#     Plot.set_xlabel(r''+CLuDict['Method']+' distance',fontsize=14)
-#     Plot.set_ylabel(r''+'data set no.',fontsize=14)
-    
-#     CLR = SCH.dendrogram(CLuZ,orientation='right',ax=Plot)
-
-#     if not newPlot:
-#         Page.toolbar.push_current()
-#         Plot.set_xlim(xylim[0])
-#         Plot.set_ylim(xylim[1])
-#         Page.toolbar.push_current()
-#         Page.ToolBarDraw()
-#     else:
-#         Page.canvas.draw()
-
-def PlotClusterXYZ(G2frame,YM,XYZ,CLuDict,Title='',PlotName=None):
-    ''' To plot cluster vectors 
+def PlotClusterXYZ(G2frame,YM,XYZ,CLuDict,Title='',PlotName='cluster'):
+    ''' To plot cluster analysis results
     :param wx.Frame G2frame: The main GSAS-II tree "window"
-    :param array whitMat: whitened data matrix
-    :param array codebook: array of cluster centers
-    
-    :param str labelX,labelY,labelZ: labels for X,Y,Z-axes
+    :param array YM: data matrix; plotted as contour
+    :param array XYZ: array of 3D PCA coordinates; plotted as 3D scatter plot
+    ;param dict CLuDict: Cluster info; may have dendogram & Kmeans results
     :param str Title: plot title
     :param str PlotName: plot tab name
@@ -11626 +11585 @@
             
     def OnPick(event):
-        ind = event.ind
-        print(CLuDict['Files'][ind[0]])
+        line = event.artist
+        ind = int(line.get_label().split('tion')[1])
+        text = 'Data selected: %s'%(CLuDict['Files'][ind])
+        G2frame.G2plotNB.status.SetStatusText(text,1)
+        print(text)
             
     Colors = ['xkcd:blue','xkcd:red','xkcd:green','xkcd:cyan', 
@@ -11642 +11604 @@
         Page.canvas.mpl_connect('motion_notify_event', OnMotion)
         Page.canvas.mpl_connect('pick_event', OnPick)
-    G2frame.G2plotNB.status.SetStatusText('',1)
     Page.Choice = None
     np.seterr(all='ignore')
@@ -11648 +11609 @@
     Imin = np.min(YM)
     Imax = np.max(YM)
+    if CLuDict['CLuZ'] is None and CLuDict['plots'] == 'Dendogram':
+        CLuDict['plots'] = 'All'
     if CLuDict['plots'] == 'Distances':
         Page.ImgObj = Plot.imshow(YM,interpolation='nearest',vmin=Imin,vmax=Imax,origin='lower')
@@ -11689 +11652 @@
         ax1.set_ylabel('Data set',fontsize=12)
         if CLuDict['codes'] is not None:
-            ax2.scatter(XYZ[0],XYZ[1],XYZ[2],color=[Colors[code] for code in CLuDict['codes']],picker=True)
+            for ixyz,xyz in enumerate(XYZ.T):
+                ax2.scatter(xyz[0],xyz[1],xyz[2],color=Colors[CLuDict['codes'][ixyz]],picker=True)
+#            ax2.scatter(XYZ[0],XYZ[1],XYZ[2],color=[Colors[code] for code in CLuDict['codes']],picker=True)
         else:
-            ax2.scatter(XYZ[0],XYZ[1],XYZ[2],color=Colors[0],picker=True)
+            for ixyz,xyz in enumerate(XYZ.T):
+                ax2.scatter(xyz[0],xyz[1],xyz[2],color=Colors[0],picker=True)
+#            ax2.scatter(XYZ[0],XYZ[1],XYZ[2],color=Colors[0],picker=True)
         ax2.set_xlabel('PCA axis-1',fontsize=12)
         ax2.set_ylabel('PCA axis-2',fontsize=12)

trunk/GSASIIseqGUI.py

@@ -1568 +1568 @@
     import scipy.cluster.hierarchy as SCH
     import scipy.cluster.vq as SCV
-    global CLuZ
-    
-        
+    import sklearn.cluster as SKC    
+        
+    SKLearnCite = '''If you use Scikit-Learn Cluster Analysis, please cite:
+    'Scikit-learn: Machine Learning in Python', Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V.,
+    Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J.,
+    Passos, A., Cournapeau, D., Brucher, M., Perrot, M. and Duchesnay, E., 
+    Journal of Machine Learning Research (2011) 12, 2825-2830.
+'''
+
     def FileSizer():
         
@@ -1828 +1834 @@
                 U,s,VT = nl.svd(YM) #s are the Eigenvalues
                 ClusData['PCA'] = s
+                s[3:] = 0.
                 S = np.diag(s)
-                XYZ = np.dot(S[:3,:3],VT[:3,:])
-                G2plt.PlotClusterXYZ(G2frame,YM,XYZ,ClusData,PlotName=ClusData['Method'],Title=ClusData['Method'])
+                XYZ = np.dot(U,np.dot(S,VT))
+                G2plt.PlotClusterXYZ(G2frame,XYZ,XYZ[:3,:],ClusData,PlotName=ClusData['Method'],Title=ClusData['Method'])
                 G2G.HorizontalLine(mainSizer,G2frame.dataWindow)
                 mainSizer.Add(wx.StaticText(G2frame.dataWindow,label='Hierarchical Cluster Analysis:'))
@@ -1845 +1852 @@
             plotSizer = wx.BoxSizer(wx.HORIZONTAL)
             plotSizer.Add(wx.StaticText(G2frame.dataWindow,label='Plot selection: '),0,WACV)
-            choice = ['All','Distances','Dendogram','3D PCA',]
+            if ClusData['CLuZ'] is None:
+                choice = ['All','Distances','3D PCA',]
+            else:
+                choice = ['All','Distances','Dendogram','3D PCA',]
             plotsel = wx.ComboBox(G2frame.dataWindow,choices=choice,style=wx.CB_READONLY|wx.CB_DROPDOWN)
             plotsel.SetValue(str(ClusData['plots']))
@@ -1851 +1861 @@
             plotSizer.Add(plotsel,0,WACV)
             mainSizer.Add(plotSizer)
+            
+        if ClusData['SKLearn']:
+            G2G.HorizontalLine(mainSizer,G2frame.dataWindow)
+            mainSizer.Add(wx.StaticText(G2frame.dataWindow,label=SKLearnCite))