def get_dunn_index(fdist, *clusters):
"""
Returns the Dunn index for the given selection of nodes.
J.C. Dunn. Well separated clusters and optimal fuzzy
partitions. 1974. J.Cybern. 4. 95-104.
"""
if len(clusters) < 2:
raise ValueError("At least 2 clusters are required")
intra_dist = []
for c in clusters:
for i in c.get_leaves():
if i is not None:
# item intraclsuterdist -> Centroid Diameter
a = fdist(i.profile, c.profile) * 2
intra_dist.append(a)
max_a = numpy.max(intra_dist)
inter_dist = []
for i, ci in enumerate(clusters):
for cj in clusters[i + 1:]:
# intracluster dist -> Centroid Linkage
b = fdist(ci.profile, cj.profile)
inter_dist.append(b)
min_b = numpy.min(inter_dist)
if max_a == 0.0:
D = 0.0
else:
D = min_b / max_a
return D
def get_dunn_index(fdist, *clusters):
"""
Returns the Dunn index for the given selection of nodes.
J.C. Dunn. Well separated clusters and optimal fuzzy
partitions. 1974. J.Cybern. 4. 95-104.
"""
if len(clusters)<2:
raise ValueError("At least 2 clusters are required")
intra_dist = []
for c in clusters:
for i in c.get_leaves():
if i is not None:
# item intraclsuterdist -> Centroid Diameter
a = fdist(i.profile, c.profile)*2
intra_dist.append(a)
max_a = numpy.max(intra_dist)
inter_dist = []
for i, ci in enumerate(clusters):
for cj in clusters[i+1:]:
# intracluster dist -> Centroid Linkage
b = fdist(ci.profile, cj.profile)
inter_dist.append(b)
min_b = numpy.min(inter_dist)
if max_a == 0.0:
D = 0.0
else:
D = min_b / max_a
return D
def heatmap(node):
square_size = 10
# Extras node info
node.collapsed = False
# Color and style
node.img_style["fgcolor"] = "#3333FF"
node.img_style["size"] = 0
ncols = node.arraytable.matrix.shape[1]
matrix_max = numpy.max(node.arraytable._matrix_max)
matrix_min = numpy.min(node.arraytable._matrix_min)
print(matrix_max)
matrix_avg = matrix_min + ((matrix_max - matrix_min) / 2)
ProfileFace = faces.ProfileFace(\
matrix_max,\
matrix_min,\
matrix_avg,\
square_size*ncols,\
square_size,\
"heatmap")
ProfileFace.ymargin = 0
if node.is_leaf():
# Set colors
faces.add_face_to_node(ProfileFace, node, 0, aligned=True)
def heatmap(node):
square_size = 10
# Extras node info
node.collapsed = False
# Color and style
node.img_style["fgcolor"] = "#3333FF"
node.img_style["size"] = 0
ncols = node.arraytable.matrix.shape[1]
matrix_max = numpy.max(node.arraytable._matrix_max)
matrix_min = numpy.min(node.arraytable._matrix_min)
print(matrix_max)
matrix_avg = matrix_min+((matrix_max-matrix_min)/2)
ProfileFace = faces.ProfileFace(\
matrix_max,\
matrix_min,\
matrix_avg,\
square_size*ncols,\
square_size,\
"heatmap")
ProfileFace.ymargin=0
if node.is_leaf():
# Set colors
faces.add_face_to_node(ProfileFace, node, 0, aligned=True)
def cluster_bars(node):
# Extras node info
node.collapsed = False
# Color and style
node.img_style["fgcolor"] = "#3333FF"
node.img_style["size"] = 4
if node.is_leaf():
matrix_max = numpy.max(node.arraytable._matrix_max)
matrix_min = numpy.min(node.arraytable._matrix_min)
matrix_avg = matrix_min + ((matrix_max - matrix_min) / 2)
ProfileFace = faces.ProfileFace(\
matrix_max,\
matrix_min,\
matrix_avg,\
200,\
40,\
"bars")
nameFace = faces.AttrFace("name", fsize=6)
faces.add_face_to_node(nameFace, node, 1, aligned=True)
faces.add_face_to_node(ProfileFace, node, 0, aligned=True)
def cluster_bars(node):
# Extras node info
node.collapsed = False
# Color and style
node.img_style["fgcolor"] = "#3333FF"
node.img_style["size"] = 4
if node.is_leaf():
matrix_max = numpy.max(node.arraytable._matrix_max)
matrix_min = numpy.min(node.arraytable._matrix_min)
matrix_avg = matrix_min+((matrix_max-matrix_min)/2)
ProfileFace = faces.ProfileFace(\
matrix_max,\
matrix_min,\
matrix_avg,\
200,\
40,\
"bars")
nameFace = faces.AttrFace("name",fsize=6 )
faces.add_face_to_node(nameFace, node, 1, aligned=True )
faces.add_face_to_node(ProfileFace, node, 0, aligned=True )
def get_min_vector(vlist):
a = numpy.array(vlist)
return numpy.min(a, 0)
def get_min_vector(vlist):
a = numpy.array(vlist)
return numpy.min(a,0)
