English

An Optimal Separation between Two Property Testing Models for Bounded Degree Directed Graphs

Data Structures and Algorithms 2023-05-23 v1

Abstract

We revisit the relation between two fundamental property testing models for bounded-degree directed graphs: the bidirectional model in which the algorithms are allowed to query both the outgoing edges and incoming edges of a vertex, and the unidirectional model in which only queries to the outgoing edges are allowed. Czumaj, Peng and Sohler [STOC 2016] showed that for directed graphs with both maximum indegree and maximum outdegree upper bounded by dd, any property that can be tested with query complexity Oε,d(1)O_{\varepsilon,d}(1) in the bidirectional model can be tested with n1Ωε,d(1)n^{1-\Omega_{\varepsilon,d}(1)} queries in the unidirectional model. In particular, if the proximity parameter ε\varepsilon approaches 00, then the query complexity of the transformed tester in the unidirectional model approaches nn. It was left open if this transformation can be further improved or there exists any property that exhibits such an extreme separation. We prove that testing subgraph-freeness in which the subgraph contains kk source components, requires Ω(n11k)\Omega(n^{1-\frac{1}{k}}) queries in the unidirectional model. This directly gives the first explicit properties that exhibit an Oε,d(1)O_{\varepsilon,d}(1) vs Ω(n1f(ε,d))\Omega(n^{1-f(\varepsilon,d)}) separation of the query complexities between the bidirectional model and unidirectional model, where f(ε,d)f(\varepsilon,d) is a function that approaches 00 as ε\varepsilon approaches 00. Furthermore, our lower bound also resolves a conjecture by Hellweg and Sohler [ESA 2012] on the query complexity of testing kk-star-freeness.

Keywords

Cite

@article{arxiv.2305.13089,
  title  = {An Optimal Separation between Two Property Testing Models for Bounded Degree Directed Graphs},
  author = {Pan Peng and Yuyang Wang},
  journal= {arXiv preprint arXiv:2305.13089},
  year   = {2023}
}

Comments

To appear in ICALP 2023

R2 v1 2026-06-28T10:41:30.711Z