Edge removal in the undirected network

Three different edge removal criteria are used in the analysis of network resilience, one of them being random edge removal.
Figure {TODO: REFERENCE EDGE UNDIRECTED CC SIZE} shows how different edge removal criteria affect the size of the largest connected
component. Interestingly, money flow criterion least deteriorates the network structure, even worse than random edge removal.
Therefore, we can assume that biggest sums of money are transferred between companies that are tightly connected with each other
through other paths of low length. Furthermore, there are also self edges in the graph that represent companies' internal fundings.
Over a period of one year, sum of such fundings can be immense, but removing a self edge does not deteriorate the network structure at all.
Consequently, with removal of edges of large money flow, average shortest path increases only slightly, even at high proportion of removed
edges, which can be seen in Figure {TODO: REFERENCE UNDIRECTED EDGE AVG PATH}. Moreover, as seen in Figure
{TODO: REFERENCE UNDIRECTED EDGE PERFORMANCE DROP},
network performance also drops only slightly in comparison with other edge removal criteria or node removal criteria. On the other hand,
betweenness edge removal performs a better job, but surprisingly, only as good as random edge removal. Only in terms of path lengths,
edge betweenness criterion produces networks with slightly longer paths than random removal at high proportion of removed edges.
This implies that there do not exist distinctive edge ``bridges'' in the network, whereas there exist distinctive node ``bridges'' as
seen with betweenness node removal. Therefore, there are some companies that act as ``bridges'' in the network, but they all have a
arge number of edges to other partners, so none of those edges is of paramount importance.