The classification theorem (A7) #
Theorem thm:classification: for a pair-source scenario with binary observations, some
finite order of the Navascués–Wolfe test (equivalently of the ancestral-independence test,
equivalently of the recursively expressible test) equals the compatible set iff every
connected component is a double-star (classification_NW_lib, classification_AI,
classification_exp), with the nontermination half nontermination_of_not_doubleStar
assembled from the cycle and five-path witnesses, transport, exhaustion, and local flips.
Everything here is proved.
AUDIT-NOTES A7, the nontermination half. A pair graph with a component that is not a double star carries, at every order, a full-support target that passes the recursively expressible test and is incompatible. Full support comes from independent local flips at every copied observation, which preserve symmetry and all AI products.
AUDIT-NOTES A7, the classification theorem for the Navascués–Wolfe hierarchy: some finite order of the hierarchy equals the compatible set exactly when every connected component of the graph is a double star.
The plain name TriangleInflation.Graph.classification_NW is reserved for the registry
statement, which PalomarSolutions/TriangleInflationClassification.lean declares and
discharges by this theorem; Comparator identifies the Challenge and the Solution by that one
name.
AUDIT-NOTES A7 for the ancestral-independence hierarchy.
AUDIT-NOTES A7 for the recursively expressible hierarchy.