def check_example_camerini1980ranking(
solver: rank.DescendSpanningArborescences,
attr: str,
attr_label: Union[str, None],
):
"""Test the example in [camerini1980ranking]_ for ranking SAs.
Args:
solver: the generator to return SAs.
attr: the edge attribute used to in determining optimality.
attr_label: the edge attribute used as unique ID for edge.
"""
assert solver.msa.size(weight=attr) == 28
res2 = next(solver)[0]
assert res2.size(weight=attr) == 26
res3 = next(solver)[0]
assert res3.size(weight=attr) == 19
res4 = next(solver)[0]
assert res4.size(weight=attr) == 17
res5 = next(solver)[0]
assert res5.size(weight=attr) == 15
if attr_label:
labels = lambda g: get_labels(g, attr_label)
assert labels(solver.msa) == {"e1", "e3", "e5"}
assert labels(res2) == {"e1", "e5", "e6"}
assert labels(res3) == {"e2", "e4", "e6"}
assert labels(res4) == {"e1", "e2", "e3"}
assert labels(res5) == {"e1", "e2", "e6"}
assert nx.is_frozen(solver.raw)
def test_ascend_sa(dg: nx.MultiDiGraph):
"""Check the generator to ascend spanning arborescences."""
solver = rank.AscendSpanningArborescences(dg,
root="b1",
attr=_ATTR,
attr_label=_ATTR_LABEL)
assert solver.msa.size(weight=_ATTR) == 15
assert (solver.msa.edges["b1", "b2"]["weight"] == 6
and solver.msa.edges["b1", "b4"]["weight"] == 1
and solver.msa.edges["b4", "b3"]["weight"] == 8)
res2 = next(solver)[0]
assert res2.size(weight=_ATTR) == 17
res3 = next(solver)[0]
assert res3.size(weight=_ATTR) == 19
res4 = next(solver)[0]
assert res4.size(weight=_ATTR) == 26
res5 = next(solver)[0]
assert res5.size(weight=_ATTR) == 28
# Check if all the SAs are correct.
labels = lambda x: get_labels(x, _ATTR_LABEL)
assert labels(res5) == {"e1", "e3", "e5"}
assert labels(res4) == {"e1", "e5", "e6"}
assert labels(res3) == {"e2", "e4", "e6"}
assert labels(res2) == {"e1", "e2", "e3"}
assert labels(solver.msa) == {"e1", "e2", "e6"}
assert nx.is_frozen(solver.raw)
def pformat(self):
"""Pretty formats your graph into a string.
This pretty formatted string representation includes many useful
details about your graph, including; name, type, frozeness, node count,
nodes, edge count, edges, graph density and graph cycles (if any).
"""
lines = []
lines.append("Name: %s" % self.name)
lines.append("Type: %s" % type(self).__name__)
lines.append("Frozen: %s" % nx.is_frozen(self))
lines.append("Nodes: %s" % self.number_of_nodes())
for n in self.nodes_iter():
lines.append(" - %s" % n)
lines.append("Edges: %s" % self.number_of_edges())
for (u, v, e_data) in self.edges_iter(data=True):
if e_data:
lines.append(" %s -> %s (%s)" % (u, v, e_data))
else:
lines.append(" %s -> %s" % (u, v))
lines.append("Density: %0.3f" % nx.density(self))
cycles = list(nx.cycles.recursive_simple_cycles(self))
lines.append("Cycles: %s" % len(cycles))
for cycle in cycles:
buf = six.StringIO()
buf.write("%s" % (cycle[0]))
for i in range(1, len(cycle)):
buf.write(" --> %s" % (cycle[i]))
buf.write(" --> %s" % (cycle[0]))
lines.append(" %s" % buf.getvalue())
return os.linesep.join(lines)
def clean_up(self):
if nx.is_frozen(self.G):
return
if self.use_virtual_nodes:
if self.verbose > 0:
print("Remove virtual nodes")
print("\nCurrent graph:")
print("Nodes: {}".format(self.G.number_of_nodes()))
print("Edges: {}".format(self.G.number_of_edges()))
self.G.remove_nodes_from(self.virtual_nodes)
self.assignments = np.delete(self.assignments, self.virtual_nodes)
self.fixed = np.delete(self.fixed, self.virtual_nodes)
if self.verbose > 0:
print("\nVirtual nodes removed:")
print("Nodes: {}".format(self.G.number_of_nodes()))
print("Edges: {}".format(self.G.number_of_edges()))
# Add partition attribute to nodes
for i in range(0, len(self.assignments)):
self.G.add_nodes_from([i], partition=str(self.assignments[i]))
# Remove original node/edge weights
for node in self.G.nodes_iter(data=True):
if 'weight_orig' in node[1]:
del node[1]['weight_orig']
for edge in self.G.edges_iter(data=True):
if 'weight_orig' in edge[2]:
del edge[2]['weight_orig']
# Freeze Graph from further modification
self.G = nx.freeze(self.G)
def _remove_alarms_of_other_projects(self, graph, current_project_id,
is_admin_project):
"""Removes wrong alarms from the graph
Removes alarms of other tenants from the graph, In case the tenant is
admin then project_id can also be None.
:type graph: NXGraph
:type current_project_id: string
:type is_admin_project: boolean
"""
alarms_to_remove = []
for alarm in graph.get_vertices(query_dict=base.ALARMS_ALL_QUERY):
if not alarm.get(VProps.PROJECT_ID, None):
cat_filter = {VProps.VITRAGE_CATEGORY: EntityCategory.RESOURCE}
resource_neighbors = \
self.entity_graph.neighbors(alarm.vertex_id,
vertex_attr_filter=cat_filter)
if len(resource_neighbors) > 0:
resource_proj_id = \
resource_neighbors[0].get(VProps.PROJECT_ID, None)
cond1 = is_admin_project and resource_proj_id and \
resource_proj_id != current_project_id
cond2 = not is_admin_project and \
(not resource_proj_id or
resource_proj_id != current_project_id)
if cond1 or cond2:
alarms_to_remove.append(alarm)
if alarms_to_remove and is_frozen(graph._g):
graph._g = graph._g.copy()
for alarm in alarms_to_remove:
graph.remove_vertex(alarm)
def pformat(graph):
"""Pretty formats your graph into a string representation that includes
details about your graph, including; name, type, frozeness, node count,
nodes, edge count, edges, graph density and graph cycles (if any).
"""
lines = []
lines.append("Name: %s" % graph.name)
lines.append("Type: %s" % type(graph).__name__)
lines.append("Frozen: %s" % nx.is_frozen(graph))
lines.append("Nodes: %s" % graph.number_of_nodes())
for n in graph.nodes_iter():
lines.append(" - %s" % n)
lines.append("Edges: %s" % graph.number_of_edges())
for (u, v, e_data) in graph.edges_iter(data=True):
if e_data:
lines.append(" %s -> %s (%s)" % (u, v, e_data))
else:
lines.append(" %s -> %s" % (u, v))
lines.append("Density: %0.3f" % nx.density(graph))
cycles = list(nx.cycles.recursive_simple_cycles(graph))
lines.append("Cycles: %s" % len(cycles))
for cycle in cycles:
buf = six.StringIO()
buf.write(str(cycle[0]))
for i in range(1, len(cycle)):
buf.write(" --> %s" % (cycle[i]))
buf.write(" --> %s" % (cycle[0]))
lines.append(" %s" % buf.getvalue())
return "\n".join(lines)
def append_loci(self, loci, **kwargs):
if nx.is_frozen(self.DAG):
raise AttributeError('The structure has been fixed')
name = loci.Name
if name in self.DAG:
if not isinstance(self.DAG.nodes[name]['loci'], ExoValueLoci):
raise KeyError('Duplicated variable name')
self.DAG.add_node(name, loci=loci, **kwargs)
self.DAG.remove_in_edges(name)
new_pa = list()
for pa in loci.Parents:
if pa not in self.DAG:
self.append_loci(ExoValueLoci(pa))
new_pa.append(pa)
self.DAG.add_edge(pa, name)
# Check acyclic or not
if not self.DAG.check_acyclic():
self.DAG.remove_node(name)
for pa in new_pa:
self.DAG.remove_node(pa)
raise AttributeError('The node causes cyclic paths')
def recurse(G, comm_limit): # RECURSIVE CALLS TO FORM COMUNITIES
if comm_limit <= 0:
return
else:
comm_limit -= 1
key, value = comparator(G)
if value == -1:
return
print '\n', "MAX_EDGE_BETWEENESS_FOUND_FOR:", key
all_comms = communities(G, key)
print "COMMUNITIES_FOUND_DUE_TO_SPLIT_THROUGH:", key, " ; SPLIT_LEVEL:", comm_limit + 1, '\n'
Graph_ar = []
das = 1
for i in all_comms:
print "COMMUNITY ", das, ' : ', i
das += 1
Graph_ar.append(G.subgraph(i))
for i in Graph_ar:
if len(i) <= 1:
continue
j = nx.Graph(i)
if nx.is_frozen(j):
print "FROZEN_ERROR"
continue
recurse(j, comm_limit)
def apply_edge_attr_filter(graph, edge_attr_filter):
edges = graph._g.edges(data=True, keys=True)
edges_to_remove = [(u, v, k) for (u, v, k, d) in edges
if not check_filter(d, edge_attr_filter)]
if edges_to_remove and is_frozen(graph._g):
graph._g = graph._g.copy()
for source, target, key in edges_to_remove:
graph._g.remove_edge(u=source, v=target, key=key)
def test_descend_sa(dg: nx.MultiDiGraph):
"""Check the generator to descend spanning arborescences.
Args:
dg: case in [camerini1980ranking]_ with 4 buses and 6 edges.
"""
solver = rank.DescendSpanningArborescences(dg,
root="b1",
attr=_ATTR,
attr_label=_ATTR_LABEL)
check_example_camerini1980ranking(solver, _ATTR, _ATTR_LABEL)
assert not nx.is_frozen(dg)
def __resGraphs(self, residualEdges):
nds = []
for e in residualEdges:
nds += list(e)
g2 = self.G.subgraph(set(nds))
assert nx.is_frozen(g2)
# To “unfreeze” a graph you must make a copy by creating a new graph object:
g2 = nx.DiGraph(g2)
edges_should_be_removed = []
for e in g2.edges:
if e not in residualEdges:
edges_should_be_removed.append(e)
g2.remove_edges_from(edges_should_be_removed)
return g2
def shave(self, seed_cpds):
'''
shave off nodes that do not connect seeds, i.e.
any node with degree = 1 and is not a seed, iteratively.
'''
if nx.is_frozen(self.graph):
gg = self.graph
self.graph = gg ## not sure why graph gets frozen but unfreeze it
nonseeds = [x for x in self.graph.nodes() if x not in seed_cpds]
excessive = [x for x in nonseeds if self.graph.degree(x) == 1]
while excessive:
for x in excessive:
self.graph.remove_node(x)
nonseeds = [x for x in self.graph.nodes() if x not in seed_cpds]
excessive = [x for x in nonseeds if self.graph.degree(x) == 1]
def _common_format(g, edge_notation):
lines = []
lines.append("Name: %s" % g.name)
lines.append("Type: %s" % type(g).__name__)
lines.append("Frozen: %s" % nx.is_frozen(g))
lines.append("Density: %0.3f" % nx.density(g))
lines.append("Nodes: %s" % g.number_of_nodes())
for n in g.nodes_iter():
lines.append(" - %s" % n)
lines.append("Edges: %s" % g.number_of_edges())
for (u, v, e_data) in g.edges_iter(data=True):
if e_data:
lines.append(" %s %s %s (%s)" % (u, edge_notation, v, e_data))
else:
lines.append(" %s %s %s" % (u, edge_notation, v))
return lines
def test_descend_sa_label_auto(dg_label: nx.DiGraph):
"""Check if the graph is labelled automatically.
Note:
``label_`` is the default label name in the algorithm, no matter
if there are existing attribute values. That is, if there are
values for the ``label_`` attribute and ``attr_label`` is set to
``label_``, those values will be updated by automatically
generated values.
"""
solver = rank.DescendSpanningArborescences(
dg_label,
root="b1",
attr=_ATTR,
attr_label="label_",
)
assert solver.msa.size(weight=_ATTR) == 28
check_example_camerini1980ranking(solver, _ATTR, None)
assert not nx.is_frozen(dg_label)
def _common_format(g, edge_notation):
lines = []
lines.append("Name: %s" % g.name)
lines.append("Type: %s" % type(g).__name__)
lines.append("Frozen: %s" % nx.is_frozen(g))
lines.append("Density: %0.3f" % nx.density(g))
lines.append("Nodes: %s" % g.number_of_nodes())
for n, n_data in g.nodes_iter(data=True):
if n_data:
lines.append(" - %s (%s)" % (n, n_data))
else:
lines.append(" - %s" % n)
lines.append("Edges: %s" % g.number_of_edges())
for (u, v, e_data) in g.edges_iter(data=True):
if e_data:
lines.append(" %s %s %s (%s)" % (u, edge_notation, v, e_data))
else:
lines.append(" %s %s %s" % (u, edge_notation, v))
return lines
def getGraphDegree(graph):
assert (nx.is_frozen(graph))
# pinNeighbors = collections.defaultdict(lambda: []) #pinId : list(pint node degrees)
# boardNeighbors = collections.defaultdict(lambda: []) #pinId : list(board node degrees)
pinOut = collections.defaultdict(lambda: [])
# boardOut = collections.defaultdict(int)
for k, node in enumerate(list(graph.nodes())):
if config.pin_token not in node: continue #only count pins
currPinList = []
currBoardList = []
for i, neighbor in enumerate(nx.all_neighbors(graph, node)):
#this line of code shouldn't ever hit, keeping it here for sanity sake
if config.board_token not in neighbor:
assert (False)
continue #only count boards
# boardNeighbors[node].append(graph.degree(pin_neighbors))
currBoardList.append(graph.degree(neighbor))
for j, pin_neighbors in enumerate(nx.all_neighbors(
graph, neighbor)):
#this condition should also never trigger
if config.pin_token not in pin_neighbors:
assert (False)
continue #only count second degree pins
currPinList.append(graph.degree(pin_neighbors))
# pinNeighbors[node].append(graph.degree(pin_neighbors))
currPinList = np.array(currPinList)
currBoardList = np.array(currBoardList)
pinOut[node] = [graph.degree(node),
currPinList.mean(), currPinList.std(),\
currBoardList.mean(), currBoardList.std()]
# if k == 10:
# break
return pinOut
def pformat(graph):
lines = []
lines.append("Name: %s" % graph.name)
lines.append("Type: %s" % type(graph).__name__)
lines.append("Frozen: %s" % nx.is_frozen(graph))
lines.append("Nodes: %s" % graph.number_of_nodes())
for n in graph.nodes_iter():
lines.append(" - %s" % n)
lines.append("Edges: %s" % graph.number_of_edges())
for (u, v, e_data) in graph.edges_iter(data=True):
reason = e_data.get('reason', '??')
lines.append(" %s -> %s (%s)" % (u, v, reason))
cycles = list(nx.cycles.recursive_simple_cycles(graph))
lines.append("Cycles: %s" % len(cycles))
for cycle in cycles:
buf = six.StringIO()
buf.write(str(cycle[0]))
for i in range(1, len(cycle)):
buf.write(" --> %s" % (cycle[i]))
buf.write(" --> %s" % (cycle[0]))
lines.append(" %s" % buf.getvalue())
return "\n".join(lines)
def __init__(self,
graph: Graph,
assignment: List[int],
components: Optional[DefaultDict[int, Component]] = None,
component_scores: Optional[Dict[int, Dict[str,
float]]] = None):
if not Chromosome.objectives:
raise ClassNotInitializedException(Chromosome)
if not is_frozen(graph):
self._graph = freeze(graph)
else:
self._graph = graph
if 'order' not in graph.graph:
# require an order on the graph for consistency
# order[i] = j implies that vertex i is located at index j in vertex_set
self._graph.graph['order'] = {
vertex: idx
for idx, vertex in enumerate(self._graph)
}
self._assignment: List[int] = assignment
self._components: Dict[int, Component] = defaultdict(set)
self._component_scores: Dict[int, Dict[str, float]] = {}
self._scores: List[float] = None
if components:
self._components = components
else:
self._rebuild_components()
if component_scores:
self._component_scores = component_scores
else:
self._compute_component_scores()
self._compute_scores()
def test_is_frozen(self):
assert_equal(nx.is_frozen(self.G), False)
G = nx.freeze(self.G)
assert_equal(G.frozen, nx.is_frozen(self.G))
assert_equal(G.frozen, True)
def is_frozen(self):
return nx.is_frozen(self.DAG)
def test_is_frozen(self):
assert not nx.is_frozen(self.G)
G = nx.freeze(self.G)
assert G.frozen == nx.is_frozen(self.G)
assert G.frozen
def reroot_neuron(x, new_root, inplace=False):
""" Reroot neuron to new root.
Parameters
----------
x : CatmaidNeuron | CatmaidNeuronList
List must contain a SINGLE neuron.
new_root : int | str
Node ID or tag of the node to reroot to.
inplace : bool, optional
If True the input neuron will be rerooted.
Returns
-------
CatmaidNeuron
Rerooted neuron. Only if ``inplace=False``.
See Also
--------
:func:`~pymaid.CatmaidNeuron.reroot`
Quick access to reroot directly from CatmaidNeuron/List
objects.
Examples
--------
>>> n = pymaid.get_neuron(16)
>>> # Reroot neuron to its soma
>>> n2 = pymaid.reroot_neuron(n, n.soma)
"""
if new_root is None:
raise ValueError('New root can not be <None>')
if isinstance(x, core.CatmaidNeuron):
pass
elif isinstance(x, core.CatmaidNeuronList):
if x.shape[0] == 1:
x = x.loc[0]
else:
raise Exception('{0} neurons provided. Please provide only '
'a single neuron!'.format(x.shape[0]))
else:
raise Exception('Unable to process data of type "{0}"'.format(type(x)))
# If new root is a tag, rather than a ID, try finding that node
if isinstance(new_root, str):
if new_root not in x.tags:
logger.error('#{}: Found no treenodes with tag {} - please double '
'check!'.format(x.skeleton_id, new_root))
return
elif len(x.tags[new_root]) > 1:
logger.error('#{}: Found multiple treenodes with tag {} - please '
'double check!'.format(x.skeleton_id, new_root))
return
else:
new_root = x.tags[new_root][0]
if not inplace:
x = x.copy()
# Skip if new root is among the existing roots
if any(x.root == new_root):
if not inplace:
return x
else:
return
if x.igraph and config.use_igraph:
# Prevent warnings in the following code - querying paths between
# unreachable nodes will otherwise generate a runtime warning
with warnings.catch_warnings():
warnings.simplefilter("ignore")
# Find paths to all roots
path = x.igraph.get_shortest_paths(x.igraph.vs.find(node_id=new_root),
[x.igraph.vs.find(node_id=r) for r in x.root])
epath = x.igraph.get_shortest_paths(x.igraph.vs.find(node_id=new_root),
[x.igraph.vs.find(node_id=r) for r in x.root],
output='epath')
# Extract paths that actually worked (i.e. within a continuous fragment)
path = [p for p in path if p][0]
epath = [p for p in epath if p][0]
edges = [(s, t) for s, t in zip(path[:-1], path[1:])]
weights = [x.igraph.es[e]['weight'] for e in epath]
# Get all weights and append inversed new weights
all_weights = x.igraph.es['weight'] + weights
# Add inverse edges: old_root->new_root
x.igraph.add_edges([(e[1], e[0]) for e in edges])
# Re-set weights
x.igraph.es['weight'] = all_weights
# Remove new_root->old_root
x.igraph.delete_edges(edges)
# Get degree of old root for later categorisation
old_root_deg = len(x.igraph.es.select(_target=path[-1]))
# Translate path indices to treenode IDs
ix2id = {ix: n for ix, n in zip(x.igraph.vs.indices,
x.igraph.vs.get_attribute_values('node_id'))}
path = [ix2id[i] for i in path]
else:
# If this NetworkX graph is just an (immutable) view, turn it into a
# full, independent graph
if float(nx.__version__) < 2.2:
if isinstance(x.graph, nx.classes.graphviews.ReadOnlyGraph):
x.graph = nx.DiGraph(x.graph)
elif hasattr(x.graph, '_NODE_OK'):
x.graph = nx.DiGraph(x.graph)
elif nx.is_frozen(x.graph):
x.graph = nx.DiGraph(x.graph)
g = x.graph
# Walk from new root to old root and remove edges along the way
parent = next(g.successors(new_root), None)
if not parent:
# new_root is already the root
return
path = [new_root]
weights = []
while parent is not None:
weights.append(g[path[-1]][parent]['weight'])
g.remove_edge(path[-1], parent)
path.append(parent)
parent = next(g.successors(parent), None)
# Invert path and add weights
new_edges = [(path[i + 1], path[i],
{'weight': weights[i]}) for i in range(len(path) - 1)]
# Add inverted path between old and new root
g.add_edges_from(new_edges)
# Get degree of old root for later categorisation
old_root_deg = g.in_degree(path[-1])
# Propagate changes in graph back to treenode table
x.nodes.set_index('treenode_id', inplace=True)
x.nodes.loc[path[1:], 'parent_id'] = path[:-1]
if old_root_deg == 1:
x.nodes.loc[path[-1], 'type'] = 'slab'
elif old_root_deg > 1:
x.nodes.loc[path[-1], 'type'] = 'branch'
else:
x.nodes.loc[path[-1], 'type'] = 'end'
x.nodes.reset_index(drop=False, inplace=True)
# Set new root's parent to None
x.nodes.parent_id = x.nodes.parent_id.astype(object)
x.nodes.loc[x.nodes.treenode_id == new_root, 'parent_id'] = None
if x.igraph and config.use_igraph:
x._clear_temp_attr(exclude=['igraph', 'classify_nodes'])
else:
x._clear_temp_attr(exclude=['graph', 'classify_nodes'])
if not inplace:
return x
else:
return
def split_into_training_test_sets(g, test_set_ratio):
print("+ Getting the gcc of the original graph.")
# Keep the original graph
train_g = g.copy()
train_g.remove_edges_from(nx.selfloop_edges(train_g)) # remove self loops
train_g = train_g.subgraph(max(nx.connected_components(train_g), key=len))
if nx.is_frozen(train_g):
train_g = nx.Graph(train_g)
print("\t- Completed!")
print("+ Relabeling.")
node2newlabel = {
node: str(nodeIdx)
for nodeIdx, node in enumerate(train_g.nodes())
}
train_g = nx.relabel_nodes(G=train_g, mapping=node2newlabel, copy=True)
print("\t- Completed!")
num_of_nodes = train_g.number_of_nodes()
nodelist = list(train_g.nodes())
print("+ Splitting into train and test sets.")
num_of_edges = train_g.number_of_edges()
edges = list(train_g.edges())
test_size = int(test_set_ratio * num_of_edges)
test_g = nx.Graph()
test_g.add_nodes_from(train_g.nodes())
count = 0
idx = 0
perm = np.arange(len(edges))
while (count < test_size and count < num_of_edges):
if count % 10000 == 0:
print("{}/{}".format(count, test_size))
# Remove the chosen edge
chosen_edge = edges[perm[idx]]
train_g.remove_edge(chosen_edge[0], chosen_edge[1])
if chosen_edge[1] in nx.connected._plain_bfs(train_g, chosen_edge[0]):
test_g.add_edge(chosen_edge[0], chosen_edge[1])
count += 1
else:
train_g.add_edge(chosen_edge[0], chosen_edge[1])
idx += 1
print("--> Completed!")
if count != test_size:
raise ValueError("Enough positive edge samples could not be found!")
# Generate the negative samples
print("\+ Generating negative samples")
count = 0
negative_samples_idx = [[] for _ in range(num_of_nodes)]
negative_samples = []
while count < 2 * test_size:
if count % 10000 == 0:
print("{}/{}".format(count, 2 * test_size))
uIdx = np.random.randint(num_of_nodes - 1)
vIdx = np.random.randint(uIdx + 1, num_of_nodes)
if vIdx not in negative_samples_idx[uIdx]:
negative_samples_idx[uIdx].append(vIdx)
u = nodelist[uIdx]
v = nodelist[vIdx]
negative_samples.append((u, v))
count += 1
train_neg_samples = negative_samples[:test_size]
test_neg_samples = negative_samples[test_size:test_size * 2]
return train_g, test_g, train_neg_samples, test_neg_samples
>>>>>>> 4c2410df82583108d45234d551042650277ca759
if np.random.rand() < test_ratio:
train_g.remove_edge(edge[0], edge[1])
else:
test_g.remove_edge(edge[0], edge[1])
# Get the greatest components and relabel
if not nx.is_connected(train_g):
print("First size of train graph: {}".format(train_g.number_of_nodes()))
train_g = train_g.subgraph(max(nx.connected_components(train_g), key=len))
print("Gcc size of train graph: {}".format(train_g.number_of_nodes()))
train_nodes = list(train_g.nodes())
node2newlabel = dict(zip(train_nodes, range(train_g.number_of_nodes())))
if nx.is_frozen(train_g):
print("Graph is frozen!")
train_g = nx.Graph(train_g)
nx.relabel_nodes(train_g, node2newlabel, copy=False)
test_g = test_g.subgraph(train_nodes)
if nx.is_frozen(test_g):
print("Graph is frozen!")
test_g = nx.Graph(test_g)
nx.relabel_nodes(test_g, node2newlabel, copy=False)
return train_g, test_g
def split(graph_path, output_folder ):
def test_is_frozen(self):
assert_equal(networkx.is_frozen(self.G), False)
G=networkx.freeze(self.G)
assert_equal(G.frozen, networkx.is_frozen(self.G))
assert_equal(G.frozen,True)
def test_is_frozen(self):
assert nx.is_frozen(self.G) == False
G = nx.freeze(self.G)
assert G.frozen == nx.is_frozen(self.G)
assert G.frozen == True
def getClusterCoef(graph, nodeDict):
assert (nx.is_frozen(graph))
#we can do this because the graph is static
neighborDict = {} #cache of all such boards
for k, node in enumerate(list(graph.nodes())):
if config.pin_token not in node: continue
boardList = set()
k_i = 0.
for i, neighbor in enumerate(nx.all_neighbors(graph, node)):
#this condition sohuldn't ever trigger
if config.board_token not in neighbor:
assert (False)
continue
boardList.add(neighbor)
k_i += graph.degree(neighbor)
if k_i == 1 or k_i == 0:
nodeDict[node].append(0.)
continue
e_i = 0. #computing e_i here
neighborList = list(nx.all_neighbors(graph, node))
# for i, neighbor_b1 in enumerate(nx.all_neighbors(graph, node)):
# for j, neighbor_b2 in enumerate(nx.all_neighbors(graph, node)):
# if neighbor_b1 == neighbor_b2: continue
for i in range(len(neighborList)):
neighbor_b1 = neighborList[i]
if neighbor_b1 in neighborDict:
b1_neighbors = neighborDict[neighbor_b1]
else:
b1_neighbors = set(nx.all_neighbors(graph, neighbor_b1))
neighborDict[neighbor_b1] = b1_neighbors
for j in range(len(neighborList[i + 1:])):
neighbor_b2 = neighborList[j]
if neighbor_b2 in neighborDict:
b2_neighbors = neighborDict[neighbor_b2]
else:
b2_neighbors = set(nx.all_neighbors(graph, neighbor_b2))
neighborDict[neighbor_b2] = b2_neighbors
#this code is SOOO slow
# e_i += len(set(list(nx.all_neighbors(graph, neighbor_b1)) + list(nx.all_neighbors(graph, neighbor_b2))))-1
union = b1_neighbors.union(b2_neighbors)
# if len(union) < 1:
# assert(len(union)>0)
e_i += len(union) - 1
# print "e_i is: ", e_i
# for j, pin_neighbors in enumerate(nx.all_neighbors(graph, neighbor)):
#condition shouldn't trigger
# if config.pin_token not in pin_neighbors:
# assert(False)
# continue
# for l, board_pin_neighbor in enumerate(nx.all_neighbors(graph, pin_neighbors)):
# if board_pin_neighbor in boardList:
# e_i += 1
clust_coef = e_i / (2. * k_i * (k_i - 1))
# if clust_coef >=1. or clust_coef<=0.:
# assert(clust_coef <=1. and clust_coef >= 0.)
nodeDict[node].append(clust_coef)
return nodeDict
