Provenance and data policy

Checked-in data

instances/tiny, small, reference-1k, and reference-4k are artificial graphs generated entirely by this tutorial. Their fixed parameters and seeds are defined in src/instance.rs; cargo run --release --locked -- --mode generate reproduces the CSV files. Nodes have no person, account, geographic, medical, or demographic meaning.

The application is framed as abstract outreach and monitoring coverage. It is not a people-targeting or political-influence application.

External graphs

The Python source tutorial used SNAP’s ego-Facebook graph. This repository does not copy, transform, or redistribute that dataset because no explicit redistribution licence was identified for the source file. A reader who has obtained an edge list under its own terms may run:

cargo run --release --locked -- \
  --mode mo --graph /local/path/to/undirected-edge-list.txt

The importer accepts two zero-based integer endpoints per non-comment line, assigns deterministic synthetic costs, and defines no native groups. Duplicate undirected edges are deduplicated. Self-loops are dropped with a counted warning, and the count is retained in run metadata. Results from such a local input are not the checked-in publication evidence.

Method provenance

The formulation was motivated by the Media.adoc, fbcover.py, and edgecover.py examples in Dietmar Wolz’s fast-cma-es repository. This port changes the application framing, replaces external data with synthetic fixtures, adds an exactly expandable native group score, separates cardinality and weighted certificates, and adds exact tiny ILPs, throughput gating, marginal-greedy comparison, reproducible artifacts, and tests.