Source Code for Module ML.Cluster.Resemblance

  1  # $Id: Resemblance.py 778 2008-07-30 09:51:55Z glandrum $ 
  2  # 
  3  # Copyright (C) 2001-2006  greg Landrum 
  4  # 
  5  #   @@ All Rights Reserved  @@ 
  6  # 
  7  """ code for dealing with resemblance (metric) matrices 
  8   
  9      Here's how the matrices are stored: 
 10   
 11       '[(0,1),(0,2),(1,2),(0,3),(1,3),(2,3)...]  (row,col), col>row' 
 12   
 13       or, alternatively the matrix can be drawn, with indices as: 
 14   
 15         || - || 0 || 1 || 3  
 16         || - || - || 2 || 4  
 17         || - || - || - || 5  
 18         || - || - || - || -  
 19   
 20       the index of a given (row,col) pair is: 
 21         '(col*(col-1))/2 + row' 
 22   
 23  """ 
 24  import numpy 
 25   
 26   
 27 -def EuclideanDistance(inData): 
 28    """returns the euclidean metricMat between the points in _inData_ 
 29   
 30      **Arguments** 
 31   
 32       - inData: a Numeric array of data points 
 33   
 34      **Returns** 
 35   
 36         a Numeric array with the metric matrix.  See the module documentation 
 37         for the format. 
 38   
 39   
 40    """ 
 41    nObjs = len(inData) 
 42    res = numpy.zeros((nObjs*(nObjs-1)/2),numpy.float) 
 43    nSoFar = 0 
 44    for col in xrange(1,nObjs): 
 45      for row in xrange(col): 
 46        t = inData[row]-inData[col] 
 47        res[nSoFar] = sum(t*t) 
 48        nSoFar += 1 
 49    return numpy.sqrt(res) 
 50   
 51 -def CalcMetricMatrix(inData,metricFunc): 
 52    """ generates a metric matrix 
 53   
 54      **Arguments** 
 55       - inData is assumed to be a list of clusters (or anything with 
 56         a GetPosition() method) 
 57   
 58       - metricFunc is the function to be used to generate the matrix 
 59   
 60   
 61      **Returns** 
 62   
 63        the metric matrix as a Numeric array 
 64   
 65    """ 
 66    nObjs = len(inData) 
 67    res = [] 
 68    inData = map(lambda x:x.GetPosition(),inData) 
 69    return metricFunc(inData) 
 70   
 71 -def FindMinValInList(mat,nObjs,minIdx=None): 
 72    """ finds the minimum value in a metricMatrix and returns it and its indices 
 73   
 74      **Arguments** 
 75   
 76       - mat: the metric matrix 
 77   
 78       - nObjs: the number of objects to be considered 
 79   
 80       - minIdx: the index of the minimum value (value, row and column still need 
 81         to be calculated 
 82   
 83      **Returns** 
 84   
 85        a 3-tuple containing: 
 86   
 87          1) the row 
 88          2) the column 
 89          3) the minimum value itself 
 90   
 91      **Notes** 
 92   
 93        -this probably ain't the speediest thing on earth 
 94   
 95    """ 
 96    assert len(mat) == nObjs*(nObjs-1)/2, 'bad matrix length in FindMinValInList' 
 97    if minIdx is None: 
 98      minIdx = numpy.argmin(mat) 
 99   
100    nSoFar = 0 
101    col = 0 
102    while nSoFar <= minIdx: 
103      col = col + 1 
104      nSoFar += col 
105   
106    row = minIdx - nSoFar + col 
107    return row,col,mat[minIdx] 
108   
109 -def ShowMetricMat(metricMat,nObjs): 
110    """ displays a metric matrix 
111   
112     **Arguments** 
113   
114      - metricMat: the matrix to be displayed 
115       
116      - nObjs: the number of objects to display 
117   
118    """ 
119    assert len(metricMat) == nObjs*(nObjs-1)/2, 'bad matrix length in FindMinValInList' 
120    for row in xrange(nObjs): 
121      for col in xrange(nObjs): 
122        if col <= row: 
123          print '   ---    ', 
124        else: 
125          print '%10.6f'%metricMat[(col*(col-1))/2+row], 
126      print 
127   
128   
129   
130  methods = [ 
131    ("Euclidean",EuclideanDistance,"Euclidean Distance"), 
132    ] 
133   
134   
135   
136  if __name__ == '__main__': 
137    m = [.1,.2,.3,.4,.5,.6,.7,.8,.9,1.0] 
138    nObjs = 5 
139    for i in range(10): 
140      print i, FindMinValInList(m,nObjs,i) 
141