Computational Supplement to Exact Values and Saturation of Graph Learning Complexity
Abstract
Documentation-only version 1.0.1 of the reproducibility software and aggregate verification records supporting Proposition 9.2 (finite verification record) of Exact Values and Saturation of Graph Learning Complexity. The manuscript was previously titled Saturation of Learning Complexity and Complete Multipartite Graphs. This version updates the software title, version-specific identifier and current-manuscript concordance while preserving the original 19-page release snapshot. All scientific code, tests, validator and aggregate outputs are byte-identical to version 1.0.0. No manuscript PDF or LaTeX source is included. The package contains an exact quotient dynamic program and a labeled evaluator for complete bipartite graphs, rational-weight and integer-parameter regression suites, complete multipartite isoperimetric checks, and an evaluator importing no quotient-solver code for all connected labeled simple graphs on two through five vertices. It includes input domains, expected outputs, a claim-to-artifact map, checksums and smoke/full reproduction instructions. The bipartite evaluators share a module and cut helper; no full implementation independence is claimed. Recorded checks cover 600 complete bipartite graphs and 113,250 quotient states, 3,456 rational-weight instances and 560 integer-parameter instances, and 3,855 weighted instances over 771 connected labeled graphs. Computations use exact integer or rational arithmetic, Python 3.10 or newer and only the standard library. The records are aggregate deterministic finite-suite summaries, not per-instance certificates or proofs of unbounded theorems. Later star, two-vertex-shore, cactus and general-subset results have deductive proofs; their later private diagnostic code and a new general-subset implementation are not included or promised.
