def mch(font_names):
print "mch of", len(font_names), font_names
distances = numpy.array([self.data[i]["distance_to"][j] for i in font_names for j in font_names if i != j])
# et for Edge Threshold
def get_graph(et):
g = networkx.Graph()
g.add_nodes_from(font_names)
edges = [(i, j) for i in font_names for j in font_names if i < j and self.data[i]["distance_to"][j] < et]
g.add_edges_from(edges)
return g
solution = [font_names] # start with one cluster and try to break it apart
quantile = 80.0 # we want to start at 40, so double it to start
while 2.5 < quantile and 1 == len(solution):
quantile = quantile * 0.5
threshold = numpy.percentile(distances, quantile)
g = get_graph(threshold)
_, solution = networkx_mcl(g) #expand_factor = <expand_factor>,
#inflate_factor = <inflate_factor>,
#max_loop = <max_loop>,
#mult_factor = <mult_factor>)
if 1 == len(solution.keys()): return solution # nothing more we can do
# if a cluster has more than 5 members, recurse down
return [c if 5 >= len(c) else mch(c) for c in solution]
def mcl_cluster(G):
M, clusters = mcl_clustering.networkx_mcl(G)
get_clusters(M)
pp.pprint(M)
pp.pprint("Found {} clusters.".format(len(clusters)))
required=True,
type=int)
parser.add_argument('-c',
'--clean',
help='clean database before executing script',
required=False,
default=False,
action='store_true')
parser.set_defaults(verbose=False)
if __name__ == '__main__':
args = parser.parse_args()
client, G, tagsToNotes = load_graph(**vars(args))
ideasCollection = client.notes.ideas
M, clusters = networkx_mcl(G)
print("Found {} clusters".format(len(clusters)))
for clique in clusters:
hashed = hash(frozenset(clique + [
args.weight,
]))
correspondingNotes = list(
set([note['_id'] for tag in clique for note in tagsToNotes[tag]]))
idea = {
'name': None,
'tags': clique,
'size': len(clique),
'notes_ids': correspondingNotes,
'notes_ids_size': len(correspondingNotes),
'algorithm': 'mcl',
