Matt Stabeler

I just realise that it’s probably doing well because its agglomerative, a bit like how the interaction networks form over time (node by node from an ego centric view) and also it’s robust to perturbations – as characterised by the network itself. This would make over great choice for an algorithm to adapt for a distributed version, where each node learns its network over time and has to account for perturbations in the graph. Might discuss this in conclusions.