Source Code for Module SloppyCell.Vandermonde.clusterScripts

 1  import scipy 
 2   
 3  # see VdmPairwise for definition of r2PC12 and dtMatPC12 
 4  # 1. make cluster instance 
 5  # PC12cluster = cluster.HierarchicalClustering(range(48), lambda x,y: getDist(x,y,r2PC12/dtMatPC12)) 
 6  # 2. set desired linake method (see cluster.py for different options) 
 7  # PC12cluster.setLinkageMethod('single') 
 8  # 3. getlevel returns the clusters of items with requested distance of one another. 
 9  #    You must run this before you can ask for the topology 
10  # PC12cluster.getlevel(0.1) 
11  # 4. get topology (formatted as a nested tuple 
12  # PC12topo = PC12cluster.topo() 
13  # 5. convert topo into a permutation list 
14  # PC12permList = FlattenIt(PC12topo,r2PC12/dtMatPC12) 
15  # 6. get permutation matrix from permutation list 
16  # PC12permMat = makePermutationMatrix(permList) 
17  # 7. permute your jacobian 
18  # PC12permJac = scipy.dot(jacPC12,PC12permMat) 
19   
20 -def getDist(ind1,ind2,distMat): 
21      """ 
22      distMat is a distance matrix.  distMat[i,j] == distance from i to j 
23      """ 
24      return distMat[ind1,ind2] 
25   
26 -def FlattenIt(x, distMat): 
27      """ 
28      Takes topo output from hierarchical clustering and orders the list at the bottom 
29      level to correspond to the optimal permutation. 
30      """ 
31      if isinstance(x, int): 
32          return [x] 
33      a = FlattenIt(x[0], distMat) 
34      b = FlattenIt(x[1], distMat) 
35  #    print a, b 
36      d00 = distMat[a[0],b[0]] 
37      d01 = distMat[a[0],b[-1]] 
38      d10 = distMat[a[-1],b[0]] 
39      d11 = distMat[a[-1],b[-1]] 
40      dMin = min(d00, d01, d10, d11) 
41      if d00 == dMin: 
42          a.reverse() 
43      elif d01 == dMin: 
44          a.reverse() 
45          b.reverse() 
46      elif d11 == dMin: 
47          b.reverse() 
48      a.extend(b) 
49  #    print a 
50      return a 
51   
52 -def makePermutationMatrix(permList): 
53      """ 
54      Takes a list defining the permutation and makes the appropriate matrix. 
55      """ 
56      permList = scipy.array(permList) 
57      n = len(permList) 
58      if 0 not in permList: 
59          permList = permList - 1 
60      permMat = scipy.zeros((n,n),'d') 
61      for ii, jj in enumerate(permList): 
62          permMat[ii,jj] = 1. 
63      return scipy.transpose(permMat) 
64   
65   
66 -def normColumns(mat1): 
67      """ 
68      Normalizes each column of a matrix.  So far just for display purposes. 
69      """ 
70      n = len(mat1[0]) 
71      mat2 = mat1.copy() 
72      for ii in range(n): 
73          mat2[:,ii] = mat2[:,ii]/scipy.sqrt(scipy.dot(mat2[:,ii],mat2[:,ii])) 
74      return mat2 
75   
76 -def colorAlignColumns(mat1): 
77      """ 
78      Multiplies columns by +/- 1 to make them have positive dot product with neighboring 
79      columns.  So far just for display purposes. 
80      """ 
81      n = len(mat1[0]) 
82      mat2 = mat1.copy() 
83      for ii in range(n-1): 
84          if scipy.dot(mat2[:,ii],mat2[:,ii+1]) < 0.: 
85              mat2[:,ii+1] = -1.*mat2[:,ii+1] 
86      return mat2 
87