Skip to content

Umbrella: Aldous–Hoover–Kallenberg for multi-sorted relational structures #103

Description

@cameronfreer

A generic AHK program for purely relational languages (no function symbols, no constants), separate from the graph (#33-closed) and directed (#84) campaigns. Existing issues cover graph and (upcoming) digraph specializations but not the generic relational-structure / arbitrary-arity program. Deliberately a NEW umbrella — do not broaden #84 until it becomes unmanageable.

Why a separate umbrella

Signature (purely relational)

structure RelSignature where
  Sort : Type
  Rel : Type
  arity : Rel → ℕ
  argSort : (R : Rel) → Fin (arity R) → Sort
  • one sort ⟹ jointly exchangeable arrays;
  • two sorts + a binary relation ⟹ separately exchangeable / bipartite arrays;
  • several relation symbols ⟹ heterogeneous random relational structures;
  • directed graphs ⟹ one binary relation, ordered arguments, diagonals permitted.

Scope assumptions (for compact Polish infinite-structure spaces): [Fintype Sort], [Countable Rel], arity finite (supplied by Nat). Nullary relations: keep 0 < arity R as an external hypothesis supplied where needed — NOT a field of RelSignature (see #110) — so nullary relations, which introduce global bits and special dissociated-representation behavior, can be added later without redesigning the signature.

Children

R0 design checkpoint (#110); R1 finite / infinite structures (carrier / coordinate / cylinder / extensionality — #104); R2 exchangeable laws (size-vector marginals, sortwise consistency, compactness extension, invariant-law equivalence — #105); R3 generic extremality characterization (exact structure laws, NOT Möbius — #106); R4 dissociated functional AHK (subset-of-tagged-vertex latents + equality patterns — #107); R5 general mixture (ergodic decomposition + functional representation — #108); R6 specialization equivalences (#109).

Interaction with the directed campaign

Adjusted near-term order: #59 → R0 → R1 → D1 (#85) as an R1 bridge → R2 → D2 (#86) as a thin R2 bridge → D3 (#87). Build the compact-product / projective-extension machinery once in R1 / R2. D3 stays concrete (five-component CAF digraphon + categorical sampler are genuinely directed-specific).

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions