Loosely connected clusters

[srid]

Right now the z-index displays totally disconnected clusters. But in practice, this is less useful.

Consider srid.ca's z-index, which has both "Life" and "Software" groups under the same cluster. They are more or less two separate clusters ... yet, because the Neuron zettel (under "Software") links to the "Use Zettelkasten for personal reflection" zettel (also linked under "Life"), neuron treats them as one cluster. This is not very useful.

The z-index should detect clusters, discarding "loose connections" like these.

What graph algorithms exist to do this?

[pbronez]

This is related to community detection in social networks. See Community Structure on Wikipedia for some details. Girvan Newman might work well for this use case, because it basically prunes away those "loose connections." I'm not aware of a Haskel implementation, but SNAP has one in cpp.

[srid]

Closing per https://neuron.zettel.page/next

Maybe related: #589