I have been attempting to replicate a random attack on an Erdos-Renyi network. The expected outcome is for the network to collapse after removing approximately 20 to 30% of the nodes. However, my results differ as the size of the giant connected component appears to decrease linearly.
import networkx as nx
import random as rd
def simulate_random_attack():
network = nx.fast_gnp_random_graph(1000, 0.01)
nodes = []
gcc = []
total_nodes = network.number_of_nodes()
removed_nodes = 0
gcc_size = len(list(max(nx.connected_components(network), key=len))) / total_nodes
gcc.append(gcc_size)
nodes.append(0.0)
while network.number_of_nodes() > 1:
node = rd.choice(list(network.nodes()))
network.remove_node(node)
removed_nodes += 1
gcc_size = len(list(max(nx.connected_components(network), key=len))) / total_nodes
gcc.append(gcc_size)
nodes.append(1 - ((total_nodes - removed_nodes) / total_nodes))
fig, ax = plt.subplots()
ax.plot(nodes, gcc)
Expected outcome:
https://i.stack.imgur.com/7A3fi.png
My results: