As described in the previous section, alternative nodes can be selected by the composition component either globally based on absolute Q-values, or locally by picking a child node from a sorted list.
In each time step t of the model we do one of the following: (i) with probability p we add a reaction node to R t -1 by at random picking a node which is not in R t -1, but is neighbour of at least one of the current nodes (as defined in the underlying BioCyc-network).
After picking a node u, the algorithm decides to include the node into the sample with probability (min(1,k_{v} k_{u}^{-alpha})).
Start out by picking a subject.
With complementary probability 1- p we pick a node in R t -1, which is a leaf, i.e. has only one link, and remove it.
In order to keep the distributed storage consistent, data storage updates are also applied by Bamboo, where a node periodically picks a random node in its leafset and synchronizes the stored data with it.
Then, backtracking from the root to the leaves, we identify the corresponding clusters by picking the first node X, for which Q(X) = Q max (X).
Although there are many previous works that generated the subgraphs by picking nodes at random (e.g., [45]), the method does not work in this experiment as it outputs sparse graphs with a large number of small connected components.
The coarsening scheme proposed by Walshaw [ 30] was to pick a random node and match it with a neighboring node with the smallest weight (defined to be 1 for each node in G0 and the number of original nodes inside a metanode in Gi for i > 0).
At each step of the traversal, we randomly pick a child node of the currently visited node.
