How would you calculate something like a Hausdorff dimension of a graph given there’s no unique embedding in Euclidean space, but we still think of (for eg) bushy vs sparse branching etc? A long skinny DAG with few branches should be lower dim than dense random graph for eg 🤔
Conversation
Replying to
Maybe average or median fan-out per node for a directed graph? Or average number of choices (also fan-out) at each node for a DAG?
Which flavor or quality are trying to capture, transferred over from the original scalar?
1
Replying to
If you squint at a (biological) tree or cauliflower it looks like a sankey diagram of an aggregate flow.
The coarse topology of that flow given a resolution level. Like a recursive approximation of min-cut-max-flow view.
DAGs are the easiest case.
Specific application: “thickness” of time history of a software project in github, given graph of forks, merges, PRs etc
1
Show replies

