def create_label_based_sbm_global_constant_nodes(graph_path,
labels,
output_path,
control_ratio,
edges_per_node,
frm,
to,
seed=0,
gen_sbm=False):
graph = get_graph(graph_path, labels, gen_sbm, seed)
label_id_to_node_id = create_community_id_to_node_id(labels)
labels_list = label_id_to_node_id.keys()
attackers = get_nodes_with_degree_percentile(graph, frm, to)
control_num = int(len(graph.nodes()) * control_ratio)
assert control_num <= len(
attackers
), f'control_num={control_num} > len(attackers)={len(attackers)}'
# print(attackers)
attackers = np.random.choice(attackers, control_num,
replace=False).tolist()
# print(attackers)
print(f'Percentage ({frm}-{to}) controlling {len(attackers)} nodes')
# print(type(edges_per_node))
rows = []
for i in range(edges_per_node):
rows = rows + attackers
cols = np.random.choice(graph.nodes(), len(attackers) * edges_per_node)
edges = [[u, v] for u, v in zip(rows, cols)]
graph.add_edges_from(edges)
nx.write_edgelist(graph, output_path)
return graph
def create_label_based_sbm_local_edges(graph_path,
labels,
output_path,
edges_to_add,
seed=0,
gen_sbm=False):
graph = get_graph(graph_path, labels, gen_sbm, seed)
label_id_to_node_id = create_community_id_to_node_id(labels)
labels_list = list(label_id_to_node_id.keys())
subg = {
c: graph.subgraph([u for u in label_id_to_node_id[c]])
for c in labels_list
}
subg_sizes = [len(subg[c].nodes()) for c in labels_list]
p = [size / len(graph.nodes()) for size in subg_sizes]
chosen_communities = np.random.choice(labels_list,
edges_to_add,
replace=True,
p=p)
rows = [np.random.choice(subg[c], 1)[0] for c in chosen_communities]
cols = [np.random.choice(subg[c], 1)[0] for c in chosen_communities]
edges = [[u, v] for u, v in zip(rows, cols)]
print(f'{sum([graph.has_edge(*edge) for edge in edges])} edges duplicated')
graph.add_edges_from(edges)
nx.write_edgelist(graph, output_path)
return graph
def create_label_based_sbm_distant_edges(graph_path,
labels,
output_path,
edges_to_add,
seed=0,
gen_sbm=False):
graph, node_mappings, reverse_node_mappings = create_graph_and_node_mappings_from_file(
graph_path)
label_id_to_node_id = create_community_id_to_node_id(labels)
block_sizes, edge_probabilities, node_lists = build_stochastic_block_matrix(
graph, node_mappings, reverse_node_mappings, labels)
# print(edge_probabilities[0])
target = [
np.argwhere(prob == np.amin(prob)).flatten().tolist()
for prob in edge_probabilities
]
labels_list = list(label_id_to_node_id.keys())
graph = get_graph(graph_path, labels, gen_sbm, seed)
subg = {
c: graph.subgraph([u for u in label_id_to_node_id[c]])
for c in labels_list
}
subg_sizes = [len(subg[c].nodes()) for c in labels_list]
p = [size / len(graph.nodes()) for size in subg_sizes]
chosen_communities = np.random.choice(labels_list,
edges_to_add,
replace=True,
p=p)
rows = [np.random.choice(subg[c], 1)[0] for c in chosen_communities]
col_targets = [np.random.choice(target[labels[u]], 1)[0] for u in rows]
# print([labels[u] for u in rows[:10]])
# print(col_targets[:10])
# exit(0)
cols = [np.random.choice(subg[c], 1)[0] for c in col_targets]
edges = [[u, v] for u, v in zip(rows, cols)]
print(f'{sum([graph.has_edge(*edge) for edge in edges])} edges duplicated')
graph.add_edges_from(edges)
nx.write_edgelist(graph, output_path)
return graph
def create_label_based_sbm_global_edges(graph_path,
labels,
output_path,
edges_to_add,
seed=0,
gen_sbm=False):
graph = get_graph(graph_path, labels, gen_sbm, seed)
label_id_to_node_id = create_community_id_to_node_id(labels)
labels_list = label_id_to_node_id.keys()
rows = np.random.choice(graph.nodes(), edges_to_add)
cols = np.random.choice(graph.nodes(), edges_to_add)
edges = [[u, v] for u, v in zip(rows, cols)]
graph.add_edges_from(edges)
nx.write_edgelist(graph, output_path)
return graph
def create_label_based_sbm_global_degree_cat(graph_path,
labels,
output_path,
edges_to_add,
frm,
to,
seed=0,
gen_sbm=False):
graph = get_graph(graph_path, labels, gen_sbm, seed)
label_id_to_node_id = create_community_id_to_node_id(labels)
labels_list = label_id_to_node_id.keys()
attach_to = get_nodes_with_degree_percentile(graph, frm, to)
print(f'Percentage ({frm}-{to}) able to link to {len(attach_to)} nodes')
rows = np.random.choice(attach_to, edges_to_add)
cols = np.random.choice(graph.nodes(), edges_to_add)
edges = [[u, v] for u, v in zip(rows, cols)]
graph.add_edges_from(edges)
nx.write_edgelist(graph, output_path)
return graph