Set-Theoretic Primitives for Relations

• Set-Theoretic Primitives for Relations are defined as subsets of Cartesian products with key operations like union, intersection, composition, and converse.
• They form the basis for relational algebras and database models by extending binary relations to n-ary and multirelational frameworks with practical computational applications.
• Advanced applications include box embeddings in machine learning and non-standard set theories that offer new perspectives on logical and relational structures.

A relation, in its mathematical sense, is fundamentally a subset of a Cartesian product of sets. The analysis and manipulation of relations—binary or n-ary—are built from a core repertoire of set-theoretic primitives. These primitives provide the technical foundation for diverse frameworks, from classical relation algebra and relational database theory to multirelational models and modern set-theoretic embeddings for machine learning. This article comprehensively synthesizes foundational primitives, associated algebras, derived operations, and notable generalizations, drawing directly on canonical arXiv sources.

1. Foundational Set-Theoretic Primitives

Let $A$, $B$, and, more generally, $A_1, \dots, A_n$ be sets.

• Set: $A = \{ x \mid P(x) \}$, defined either extensionally or intensionally.
• Cartesian Product: $$A \times B = \{ (a, b) \mid a \in A \wedge b \in B \}$$.
• Binary Relation: $R \subseteq A \times B$.
• n-ary Relation: $R \subseteq A_1 \times \dots \times A_n$.

For a binary relation $R \subseteq A \times B$:

• Domain: $$\operatorname{dom}(R) = \{ a \in A \mid \exists b \in B,\, (a, b) \in R \}$$.
• Range: $$\operatorname{ran}(R) = \{ b \in B \mid \exists a \in A,\, (a, b) \in R \}$$.
• Inverse/Converse: $B$0.
• Composition: If $B$1, then $B$2.

These definitions generalize to higher arities by extending the product and tuple structure accordingly. The entire relational data model, including relational algebra, is directly grounded in manipulation of these set-theoretic objects via comprehensions and Boolean operations [0607039].

2. Primitive Operations and Relation Algebras

The core set-theoretic operations on relations, especially binary ones, are:

• Union: $B$3.
• Intersection: $B$4.
• Complement (relative to universe $B$5): $B$6.
• Identity Relation: $B$7.
• Diversity Relation: $B$8.
• Universal Relation: $B$9.
• Empty Relation: $A_1, \dots, A_n$0.
• Converse: $A_1, \dots, A_n$1.
• Composition: $A_1, \dots, A_n$2.

Relation algebras, including those underlying Belnap, Sugihara, Meyer, and Church logics, are structured as closure systems under these primitives, with derived operations like residual (implication), converse-complement (De Morgan negation), and relativized variants. All basic Boolean and algebraic identities apply, such as associativity, commutativity, distributivity, absorption, and De Morgan dualities. These laws yield a rich algebraic infrastructure for classical and substructural logics as reducts or subreducts of representable relation algebras (Kramer et al., 2019).

3. Primitives for n-ary Relations and Relational-Algebraic Operations

In the context of n-ary relations—essential for relational databases—operations extend via set comprehensions over tuples, often represented as attribute-indexed functions. Principal operators include:

• Union, Intersection, Difference: Defined tuple-wise.
• Cartesian Product: Produces tuples by combining attributes from two relations.
• Projection: $A_1, \dots, A_n$3 for a set of attributes $A_1, \dots, A_n$4.
• Selection: $A_1, \dots, A_n$5 for a predicate $A_1, \dots, A_n$6.
• Renaming: Via bijections on attribute indices.
• Natural Join: $A_1, \dots, A_n$7 (for $A_1, \dots, A_n$8 and $A_1, \dots, A_n$9).

Closure under these operations is immediate from the set-theoretic underpinnings. All such operations reduce to explicit set comprehensions combined with existential quantification over product domains [0607039].

4. Multirelational Generalizations and Additional Primitives

Multirelations generalize binary relations by allowing an element to be associated with a subset rather than a single element: $A = \{ x \mid P(x) \}$0. Key primitives adapt to this broader setting:

• Union/Intersection: Pointwise on $A = \{ x \mid P(x) \}$1.
• Sequential Composition: $A = \{ x \mid P(x) \}$2.
• Parallel Composition: $A = \{ x \mid P(x) \}$3.
• Iteration (Kleene Star, $A = \{ x \mid P(x) \}$4-iteration): Defined as least/greatest fixpoints over this composition.

When each output subset is restricted to a singleton, these operations collapse to ordinary relational operations, with parallel composition reducing to intersection (Furusawa et al., 2015). The multirelational framework exposes new algebraic layers (e.g., parallel monoid, c-lattice, c-quantale) relevant for concurrency and dynamic logics.

5. Unary Relation Operators: 16-Bit Primitives

A complete Boolean structure of unary operations on homogeneous binary relations arises from all Boolean combinations of $A = \{ x \mid P(x) \}$5 and $A = \{ x \mid P(x) \}$6. There are 16 such operators, indexed by bit patterns (0…F), including:

Hex	Mnemonic	Definition
0	empty	$A = \{ x \mid P(x) \}$7
1	symmetric kernel	$A = \{ x \mid P(x) \}$8
2	asymmetric kernel	$A = \{ x \mid P(x) \}$9
3	identity (no change)	$$A \times B = \{ (a, b) \mid a \in A \wedge b \in B \}$$0
5	converse	$$A \times B = \{ (a, b) \mid a \in A \wedge b \in B \}$$1
6	symmetric difference	$$A \times B = \{ (a, b) \mid a \in A \wedge b \in B \}$$2
7	symmetric closure	$$A \times B = \{ (a, b) \mid a \in A \wedge b \in B \}$$3
C	complement	$$A \times B = \{ (a, b) \mid a \in A \wedge b \in B \}$$4
F	universal	$$A \times B = \{ (a, b) \mid a \in A \wedge b \in B \}$$5

Composition yields a closed $$A \times B = \{ (a, b) \mid a \in A \wedge b \in B \}$$6 operator table. The operator lattice is a Boolean algebra, and all basic pointwise logical junctors (conjunction, disjunction, negation) distribute through. Lifting relational properties (e.g., reflexivity, symmetry) through these operators generates 81 distinct property-lifts (from 384 possibilities), with a minimal axiom basis of 124 implication laws among them, as detailed in the algebra of lifted properties (Burghardt, 2021).

6. Relational Primitives in Non-Standard Set Theory: Relations-Language Mathematics

The relations-language (RL) approach reconstructs set-theoretic primitives by taking "relation between points" as the exclusive primitive. Here, "sets" become classifications: for a reflexive relation $$A \times B = \{ (a, b) \mid a \in A \wedge b \in B \}$$7, $$A \times B = \{ (a, b) \mid a \in A \wedge b \in B \}$$8.

• Domain/Range: Defined via the truth-domain $$A \times B = \{ (a, b) \mid a \in A \wedge b \in B \}$$9 and $R \subseteq A \times B$0
• Operations: Pointwise for union, intersection, and complement.
• Composition: $R \subseteq A \times B$1.
• Ordered Pair and Cartesian Product: Encoded via unique relations $R \subseteq A \times B$2 or emulated Kuratowski pairs.

In RL, ontological distinctions of "empty set" and "singleton" disappear, with no "set of all subsets" and only relations mediating membership. The result is a set theory without Cantor's theorem, classical paradoxes, or ZF's cumulative hierarchy, matching ZF only in "ordinary" mathematics construction but fundamentally altering the primitive conceptual base (Mitroshin, 2012).

7. Set-Theoretic Primitives in Machine Learning: Box Embeddings

Modern representation learning applies set-theoretic primitives directly to the modeling of item-attribute (or entity-relation) structures. Box embeddings define each item or attribute as a box $R \subseteq A \times B$3 in $R \subseteq A \times B$4:

• Box Definition: $R \subseteq A \times B$5.
• Intersection: $R \subseteq A \times B$6.
• Union: Not box-closed; approximable by bounding box.
• Negation/Complement: Nonconvex, handled by inclusion–exclusion.

Membership is scored as $R \subseteq A \times B$7, with set-theoretic Boolean compositions computed via recursively defined intersection and inclusion–exclusion for difference. For compositional attribute queries (e.g., $R \subseteq A \times B$8), empirical results show substantial accuracy improvements over vector embeddings, particularly for intersection and negation queries. Unions remain challenging due to lack of closure in this geometric representation (Dasgupta et al., 2023).

---

The entire modern landscape of relations, their algebras, database theory, logical calculi, and set-theoretic embeddings is underpinned by these core primitives: set inclusion, Cartesian product (and its generalizations), pointwise Boolean operations, projection, composition, and identity. Each extension, whether in multirelational logic, unary operator algebras, or geometric embeddings, leverages and further refines these foundational constructs.