def identify_cluster(self, tree):
gn = tree
can_cluster = []
final_cluster = []
# print [node.name for node in PreOrderIter(gn)]
for pn in gn.children:
# print "hehe " + pn.name
first_cluster = Cluster(pn.name, pn.doc)
can_cluster.append(first_cluster)
# print "lol " + pn.name
for cn in pn.children:
next_cluster = Cluster(cn.parent.name + " " + cn.name, cn.doc)
temp = next_cluster
can_cluster.append(next_cluster)
for ccn in PreOrderIter(cn,
filter_=lambda n: n.name != cn.name):
new_cluster = Cluster(next_cluster.label + " " + ccn.name,
ccn.doc)
can_cluster.append(new_cluster)
if ccn.is_leaf:
next_cluster = temp
else:
next_cluster = new_cluster
# compute clus sim
self.compute_clus_sim(can_cluster)
final_cluster += can_cluster
can_cluster = []
# filter useless cluster
for cluster in reversed(final_cluster):
if not cluster.documents:
final_cluster.remove(cluster)
else:
for cluster2 in filter(lambda x: x != cluster, final_cluster):
if len(
list(
set(cluster.documents)
& set(cluster2.documents))) >= 1:
if len(cluster.documents) <= len(cluster2.documents):
final_cluster.remove(cluster)
break
# for i in range(0, len(final_cluster)):
# print("[%s: %s]" % (final_cluster[i].label, "; ".join([cd.title for cd in final_cluster[i].documents])))
del can_cluster
return final_cluster
def get_wf():
wf = pe.Workflow(name="main_workflow")
wf.base_dir = os.path.join(workingdir, "clustering_pipeline")
wf.config['execution']['crashdump_dir'] = wf.base_dir + "/crash_files"
##Infosource##
fs_infosource = pe.Node(util.IdentityInterface(fields=['fs']),
name="fs_infosource")
fs_infosource.iterables = ('fs', fsaverage)
hemi_infosource = pe.Node(util.IdentityInterface(fields=['hemi']),
name="hemi_infosource")
hemi_infosource.iterables = ('hemi', hemispheres)
cluster_infosource = pe.Node(util.IdentityInterface(fields=['cluster']),
name="cluster_infosource")
cluster_infosource.iterables = ('cluster', cluster_types)
n_clusters_infosource = pe.Node(
util.IdentityInterface(fields=['n_clusters']),
name="n_clusters_infosource")
n_clusters_infosource.iterables = ('n_clusters', n_clusters)
##Datagrabber##
datagrabber = pe.Node(nio.DataGrabber(
infields=['fs', 'hemi'],
outfields=['simmatrix', 'maskindex', 'targetmask']),
name="datagrabber")
datagrabber.inputs.base_directory = '/'
datagrabber.inputs.template = '*'
datagrabber.inputs.field_template = clustering_dg_template
datagrabber.inputs.template_args = clustering_dg_args
datagrabber.inputs.sort_filelist = True
wf.connect(fs_infosource, 'fs', datagrabber, 'fs')
wf.connect(hemi_infosource, 'hemi', datagrabber, 'hemi')
##clustering##
clustering = pe.Node(Cluster(), name='clustering')
clustering.inputs.epsilon = epsilon
wf.connect(hemi_infosource, 'hemi', clustering, 'hemi')
wf.connect(cluster_infosource, 'cluster', clustering, 'cluster_type')
wf.connect(n_clusters_infosource, 'n_clusters', clustering, 'n_clusters')
wf.connect(datagrabber, 'simmatrix', clustering, 'in_File')
##reinflate to surface indices##
clustermap = pe.Node(ClusterMap(), name='clustermap')
wf.connect(clustering, 'out_File', clustermap, 'clusteredfile')
wf.connect(datagrabber, 'maskindex', clustermap, 'indicesfile')
wf.connect(datagrabber, 'targetmask', clustermap, 'maskfile')
##Datasink##
ds = pe.Node(nio.DataSink(), name="datasink")
ds.inputs.base_directory = clusterdir
wf.connect(clustermap, 'clustermapfile', ds, 'clustered')
wf.connect(clustermap, 'clustermaptext', ds, 'clustered.@1')
wf.write_graph()
return wf
def computeClusters(self, symbolClasses):
for cl in symbolClasses:
cl.clusters = []
distortedClassesOfSingleClass = cl.learning_set[:]
centroids, labels = self.__computeClusters(
distortedClassesOfSingleClass)
# distinguish k clusters
for j in range(0, global_v.K):
# points of this cluster
points = []
for c in range(0, len(distortedClassesOfSingleClass)):
if labels[c] == j:
points.append(distortedClassesOfSingleClass[c].
characteristicsValues)
cluster = Cluster(centroids[j], points, cl.name, j)
cl.clusters.append(cluster)
def compute(k, training_set):
clusters = []
centroids, labels = __computeClusters(k, training_set)
# distinguish k clusters
for j in range(0,k):
# points of this cluster
points = []
for c in range(0, len(training_set)):
if labels[c] == j:
points.append(training_set[c].characteristicsValues)
cluster = Cluster(centroids[j], points, training_set[c].name, j,
give_info = False,
do_ellipsoid=False,
do_cuboid=False)
clusters.append(cluster)
return clusters
def get_wf():
wf = pe.Workflow(name="main_workflow")
wf.base_dir = os.path.join(workingdir, "consensus_pipeline")
wf.config['execution']['crashdump_dir'] = wf.base_dir + "/crash_files"
##Infosource##
#session_infosource = pe.Node(util.IdentityInterface(fields=['session']), name="session_infosource")
#session_infosource.iterables = ('session', analysis_sessions)
hemi_infosource = pe.Node(util.IdentityInterface(fields=['hemi']),
name="hemi_infosource")
hemi_infosource.iterables = ('hemi', hemispheres)
sim_infosource = pe.Node(util.IdentityInterface(fields=['sim']),
name="sim_infosource")
sim_infosource.iterables = ('sim', similarity_types)
cluster_infosource = pe.Node(util.IdentityInterface(fields=['cluster']),
name="cluster_infosource")
cluster_infosource.iterables = ('cluster', cluster_types)
n_clusters_infosource = pe.Node(
util.IdentityInterface(fields=['n_clusters']),
name="n_clusters_infosource")
n_clusters_infosource.iterables = ('n_clusters', n_clusters)
##Datagrabber for subjects##
dg_subjects = pe.Node(nio.DataGrabber(
infields=['hemi', 'cluster', 'sim', 'n_clusters'],
outfields=['all_subjects', 'maskindex', 'targetmask']),
name="dg_subjects")
dg_subjects.inputs.base_directory = '/'
dg_subjects.inputs.template = '*'
dg_subjects.inputs.field_template = consensus_dg_template
dg_subjects.inputs.template_args = consensus_dg_args
dg_subjects.inputs.sort_filelist = True
#wf.connect(session_infosource, 'session', dg_subjects, 'session')
wf.connect(hemi_infosource, 'hemi', dg_subjects, 'hemi')
wf.connect(cluster_infosource, 'cluster', dg_subjects, 'cluster')
wf.connect(sim_infosource, 'sim', dg_subjects, 'sim')
wf.connect(n_clusters_infosource, 'n_clusters', dg_subjects, 'n_clusters')
##Consensus between subjects##
intersubject = pe.Node(Consensus(), name='intersubject')
wf.connect(dg_subjects, 'all_subjects', intersubject, 'in_Files')
wf.connect(dg_subjects, 'targetmask', intersubject, 'maskfile')
##Cluster the Consensus Matrix##
intersubject_cluster = pe.Node(Cluster(), name='intersubject_cluster')
wf.connect(intersubject, 'consensus_mat', intersubject_cluster, 'in_File')
wf.connect(hemi_infosource, 'hemi', intersubject_cluster, 'hemi')
wf.connect(cluster_infosource, 'cluster', intersubject_cluster,
'cluster_type')
wf.connect(n_clusters_infosource, 'n_clusters', intersubject_cluster,
'n_clusters')
##Decompress##
clustermap = pe.Node(ClusterMap(), name='clustermap')
wf.connect(intersubject_cluster, 'out_File', clustermap, 'clusteredfile')
wf.connect(dg_subjects, 'maskindex', clustermap, 'indicesfile')
wf.connect(dg_subjects, 'targetmask', clustermap, 'maskfile')
##Datasink##
ds = pe.Node(nio.DataSink(), name="datasink")
ds.inputs.base_directory = consensusdir
wf.connect(intersubject, 'variation_mat', ds, 'variation_mat')
wf.connect(intersubject, 'consensus_mat', ds, 'consensus_not_clustered')
#wf.connect(intersession_cluster, 'out_File', ds, 'consensus_intersession')
wf.connect(clustermap, 'clustermapfile', ds, 'consensus_clustered')
wf.write_graph()
return wf