English

On the complexity of symmetric vs. functional PCSPs

Computational Complexity 2024-08-19 v3 Discrete Mathematics Data Structures and Algorithms

Abstract

The complexity of the promise constraint satisfaction problem PCSP(A,B)\operatorname{PCSP}(\mathbf{A},\mathbf{B}) is largely unknown, even for symmetric A\mathbf{A} and B\mathbf{B}, except for the case when A\mathbf{A} and B\mathbf{B} are Boolean. First, we establish a dichotomy for PCSP(A,B)\operatorname{PCSP}(\mathbf{A},\mathbf{B}) where A,B\mathbf{A}, \mathbf{B} are symmetric, B\mathbf{B} is functional (i.e. any r1r-1 elements of an rr-ary tuple uniquely determines the last one), and (A,B)(\mathbf{A},\mathbf{B}) satisfies technical conditions we introduce called dependency and additivity. This result implies a dichotomy for PCSP(A,B)\operatorname{PCSP}(\mathbf{A},\mathbf{B}) with A,B\mathbf{A},\mathbf{B} symmetric and B\mathbf{B} functional if (i) A\mathbf{A} is Boolean, or (ii) A\mathbf{A} is a hypergraph of a small uniformity, or (iii) A\mathbf{A} has a relation RAR^{\mathbf{A}} of arity at least 3 such that the hypergraph diameter of (A,RA)(A, R^{\mathbf{A}}) is at most 1. Second, we show that for PCSP(A,B)\operatorname{PCSP}(\mathbf{A},\mathbf{B}), where A\mathbf{A} and B\mathbf{B} contain a single relation, A\mathbf{A} satisfies a technical condition called balancedness, and B\mathbf{B} is arbitrary, the combined basic linear programming relaxation (BLP) and the affine integer programming relaxation (AIP) is no more powerful than the (in general strictly weaker) AIP relaxation. Balanced A\mathbf{A} include symmetric A\mathbf{A} or, more generally, A\mathbf{A} preserved by a transitive permutation group.

Keywords

Cite

@article{arxiv.2210.03343,
  title  = {On the complexity of symmetric vs. functional PCSPs},
  author = {Tamio-Vesa Nakajima and Stanislav Živný},
  journal= {arXiv preprint arXiv:2210.03343},
  year   = {2024}
}

Comments

Full version (with stronger results) of a LICS'23 paper

R2 v1 2026-06-28T02:58:50.916Z