Imagine a border river that separates two countries. On one bank live one group of people, on the other, another. The boundary between them is not always obvious: in some places the river is wide and calm, in others narrow and turbulent. But it is exactly where the current changes abruptly that the border becomes most noticeable. The SP-MGM algorithm looks for such natural gaps in data to understand where one group ends and another begins.
The method rests on the idea that every graph has its own natural oscillations, which describe how its vertices are connected. These oscillations can be thought of as waves spreading through the network. If two parts of a graph are weakly linked, a wave will sharply change its amplitude at the boundary between them. The algorithm finds the largest amplitude change and draws the boundary right there. This approach allows the number of clusters to be determined automatically, without setting it in advance.
Tests on standard graphs showed that SP-MGM often outperforms popular algorithms, especially when the network is small and groups are well separated. At the same time, it almost always yields one more cluster than the number of eigenmodes. This surprising property suggests that gaps in the eigenvectors indeed reflect the graph's structure.
Applying the algorithm to sea surface temperature data in the Niño 3.4 region identified seven clusters, which correspond to different phases of El Niño and La Niña. Interestingly, these clusters are not determined solely by the ONI index, meaning the method captures subtle differences in the "flavors" of this climate phenomenon. Such clustering could help better understand how different types of El Niño affect weather in different regions.
In summary, the SP-MGM algorithm offers a simple and elegant way to find hidden boundaries in data. It relies on the graph's internal structure rather than external parameters, and it can be applied across a wide range of fields, from social network analysis to climate regime classification. Returning to the river image: the algorithm finds the places where the current changes, and that reveals where one country ends and another begins.
The SP-MGM algorithm almost always outputs exactly m+1 clusters for m eigenmodes, although theoretically there could be up to $2^m$. This points to a deep connection between gaps in eigenvectors and graph structure.
The idea of finding hidden structures by gaps in data recalls the analysis of 'patterns' in William Gibson's novel 'Pattern Recognition', where the protagonist intuitively finds anomalies in video streams. The algorithm formalizes such a search for graphs.