graph

Node importance on a link graph

articles is a directed, weighted link graph among ~330 Wikipedia articles (src dst weight per line, weight = link count) — century pages, historical topics, and related categories. It is small enough to eyeball and lumpy enough that the rankings genuinely disagree.

./run.sh runs every graph algorithm. How to read the outputs:

• page-rank — global importance under the random-surfer model; --sink-dispersion picks the dangling-page treatment (uniform is the textbook choice, reverse/all are propagon traditions, none lets mass drain).
• hits — two scores per node: authorities (pointed at by good hubs) then hubs (pointing at good authorities), blank-line separated.
• katz-centrality — all incoming walks, discounted by --alpha; keep α below 1/λ_max or the run aborts with advice (this graph's weights are large — hence --alpha 0.05).
• degree — the counting baseline. If page-rank's order barely differs from in-degree's, the link structure isn't adding much signal.
• k-core — coreness per node: which articles sit in the densely interlinked center vs. the periphery (direction/weights ignored).
• birank — two-sided scores treating the edges as a bipartite src↔dst interaction matrix.
• components — utility: splits the graph into connected components, one edge file each.

Provenance

articles ships from propagon v1's example set (extracted from Wikipedia link data); kept verbatim so old and new outputs stay comparable.