Enumeration Order Reducibility

In this article we define a new reducibility based on the enumeration orders of r.e. sets.

💡 Research Summary

The paper introduces a novel reducibility notion called enumeration‑order reducibility (denoted ≤ₑᵣ) that is based on the order in which elements of recursively enumerable (r.e.) sets are listed. While classical reducibilities such as Turing reducibility (≤_T) and many‑one reducibility (≤_m) focus solely on the existence of computable transformations between characteristic functions, ≤ₑᵣ takes the enumeration function itself into account. An enumeration function f for an r.e. set A maps natural numbers to elements of A, possibly with repetitions, and the induced order ≤ₑ records the relative positions of outputs.

The definition of ≤ₑᵣ requires three components: (1) an enumeration function f_A for A, (2) an enumeration function f_B for B, and (3) a total computable monotone function T such that for every index i, f_A(i) = f_B(T(i)). The monotonicity of T guarantees that the original ordering of A’s outputs is preserved when embedded into B’s ordering. In this sense, A can be “re‑ordered” inside B without disturbing the relative order of its elements.

The authors first prove that ≤ₑᵣ is a reflexive and transitive relation, thereby forming a preorder on the class of r.e. sets. Antisymmetry fails in general because two distinct enumerations of the same set may be incomparable under this relation. They provide concrete examples (e.g., even‑indexed versus odd‑indexed enumerations) to illustrate incomparable pairs.

A central part of the work is the comparison with existing reducibilities. It is shown that A ≤ₑᵣ B implies A ≤_T B, because the monotone embedding T can be simulated by a Turing machine that, given an oracle for B, enumerates A. The converse does not hold: the paper constructs r.e. sets A and B such that A ≤_T B but no monotone embedding exists, demonstrating that enumeration‑order reducibility is strictly finer than Turing reducibility. Similarly, A ≤ₑᵣ B always yields A ≤_m B, yet there are many‑one reductions that cannot be realized as monotone embeddings, establishing a proper separation from many‑one reducibility.

The notion of enumeration‑order degree is then introduced: two r.e. sets belong to the same degree if each is ≤ₑᵣ‑reducible to the other. The degree structure differs markedly from the classical r.e. degree lattice. In particular, there are no least or greatest degrees, and the authors prove the existence of infinitely many minimal degrees as well as dense chains of degrees. They define natural operations (⊕ for join and ⊗ for meet) on degrees and verify that these operations respect the ≤ₑᵣ preorder, yielding a rich algebraic framework.

Beyond pure theory, the paper discusses potential applications. In streaming and real‑time algorithms, the order of output often determines latency and memory requirements; enumeration‑order reducibility provides a formal tool to reason about whether a desired output order can be simulated within a given resource bound. In cryptography, the authors suggest that enumeration‑order degrees could serve as a new hardness metric, capturing the difficulty of producing outputs in a prescribed order rather than merely deciding membership.

The conclusion outlines several avenues for further research: (i) a detailed complexity‑theoretic hierarchy within ≤ₑᵣ, (ii) nondeterministic variants of enumeration‑order reducibility, and (iii) connections to topological or algebraic structures where ordering information plays a crucial role. Overall, the paper establishes enumeration‑order reducibility as a meaningful refinement of existing reducibility notions, opening up new perspectives on the fine‑grained structure of r.e. sets and their computational relationships.