def merge_clusters(chain, clusters, Dmax=0, min_size=0):
"""
Due to the problem of chaining, some chains might overlap each other
these need to be merged
"""
eclusters = make_range(clusters, extend=Dmax)
mergeables = get_2D_overlap(chain, eclusters)
merged_chain = []
for mergeable in mergeables:
merged_mother = min(mergeable)
g = (clusters[x] for x in mergeable)
merged_cluster = itertools.chain(*g)
merged_cluster = list(set(merged_cluster))
clusters[merged_mother] = merged_cluster
if len(merged_cluster) >= min_size:
merged_chain.append(merged_mother)
# maintain the x-sort
[cluster.sort() for cluster in clusters]
print >> sys.stderr, "merging... (%d->%d)" % (len(chain),
len(merged_chain))
return merged_chain
def merge_clusters(chain, clusters, Dmax=0, min_size=0):
"""
Due to the problem of chaining, some chains might overlap each other
these need to be merged
"""
eclusters = make_range(clusters, extend=Dmax)
mergeables = get_2D_overlap(chain, eclusters)
merged_chain = []
for mergeable in mergeables:
merged_mother = min(mergeable)
g = (clusters[x] for x in mergeable)
merged_cluster = itertools.chain(*g)
merged_cluster = list(set(merged_cluster))
clusters[merged_mother] = merged_cluster
if len(merged_cluster) >= min_size:
merged_chain.append(merged_mother)
# maintain the x-sort
[cluster.sort() for cluster in clusters]
print >>sys.stderr, "merging... (%d->%d)" % (len(chain), len(merged_chain))
return merged_chain
def get_constraints(clusters, quota=(1, 1), Nmax=0):
"""
Check pairwise cluster comparison, if they overlap then mark edge as conflict
"""
qa, qb = quota
eclusters = make_range(clusters, extend=-Nmax)
# (1-based index, cluster score)
nodes = [(i + 1, c[-1]) for i, c in enumerate(eclusters)]
eclusters_x, eclusters_y, scores = zip(*eclusters)
# represents the contraints over x-axis and y-axis
constraints_x = get_1D_overlap(eclusters_x, qa)
constraints_y = get_1D_overlap(eclusters_y, qb)
return nodes, constraints_x, constraints_y
def get_constraints(clusters, quota=(1,1), Nmax=0):
"""
Check pairwise cluster comparison, if they overlap then mark edge as conflict
"""
qa, qb = quota
eclusters = make_range(clusters, extend=-Nmax)
# (1-based index, cluster score)
nodes = [(i+1, c[-1]) for i, c in enumerate(eclusters)]
eclusters_x, eclusters_y, scores = zip(*eclusters)
# represents the contraints over x-axis and y-axis
constraints_x = get_1D_overlap(eclusters_x, qa)
constraints_y = get_1D_overlap(eclusters_y, qb)
return nodes, constraints_x, constraints_y
