def ring_local_stitching(samples: List[nx.MultiDiGraph], bridges_amount: int):
samples_amount = len(samples)
if samples_amount < 2:
return
tails = [[choices(list(sample.nodes), k=bridges_amount) for _ in range(2)]
for sample in samples]
heads = [[choices(list(sample.nodes), k=bridges_amount) for _ in range(2)]
for sample in samples]
for i, sample in enumerate(samples):
j = 0 if i == len(samples) - 1 else i + 1
if i != j:
sample.add_edges_from(zip(tails[i][0], heads[j][1]))
def all_to_all_local_stitching(samples: List[nx.MultiDiGraph],
bridges_amount: int):
samples_amount = len(samples)
if samples_amount < 2:
return
tails = [[
choices(list(sample.nodes), k=bridges_amount)
for _ in range(len(samples))
] for sample in samples]
heads = [[
choices(list(sample.nodes), k=bridges_amount)
for _ in range(len(samples))
] for sample in samples]
for i, sample in enumerate(samples):
for j in range(len(samples)):
if i != j:
sample.add_edges_from(zip(tails[i][j], heads[j][i]))
def all_to_all_distributed_stitching(samples: List[nx.MultiDiGraph],
bridges_amount: int):
raw_samples = []
for sample in samples:
raw_samples.extend(sample.nodes)
random.shuffle(raw_samples)
my_remote_heads = [
choices(raw_samples, k=bridges_amount) for _ in range(mpi.size)
]
tails = [choices(raw_samples, k=bridges_amount) for _ in range(mpi.size)]
remote_heads = mpi.comm.alltoall(my_remote_heads)
for i in range(mpi.size):
if i != mpi.rank:
bridges_amount_i = min(len(remote_heads[i]), len(tails[i]))
samples[0].add_edges_from(
zip(remote_heads[i][:bridges_amount_i],
tails[i][:bridges_amount_i]))
def local_edge_sampling(
candidate_edges: List[Vertex],
nodes_amount: int) -> Tuple[nx.MultiDiGraph, List[Vertex]]:
subgraph = nx.MultiDiGraph()
nodes_sampled = 0
while nodes_sampled < nodes_amount:
subgraph.add_edges_from(
choices(candidate_edges,
k=ceil((nodes_amount - nodes_sampled) / 2)))
nodes_sampled = subgraph.number_of_nodes()
candidate_edges = list(
filter(lambda e: not subgraph.has_edge(*e), candidate_edges))
return subgraph, candidate_edges