Group Testing in Arbitrary Hypergraphs and Related Combinatorial Structures

TL;DR

This paper generalizes group testing to hypergraphs, developing new algorithms and combinatorial structures to improve testing bounds by leveraging hypergraph-specific characteristics.

Contribution

It introduces new combinatorial structures and algorithms for hypergraph-based group testing, extending classical concepts to more complex contamination models.

Findings

Derived bounds on the number of tests for hypergraph group testing algorithms.

Established a correspondence between hypergraph testing algorithms and new combinatorial structures.

Improved upper bounds by leveraging hypergraph characteristics.

Abstract

We consider a generalization of group testing where the potentially contaminated sets are the members of a given hypergraph ${\cal F}=(V,E)$. This generalization finds application in contexts where contaminations can be conditioned by some kinds of social and geographical clusterings. We study non-adaptive algorithms, two-stage algorithms, and three-stage algorithms. Non-adaptive group testing algorithms are algorithms in which all tests are decided beforehand and therefore can be performed in parallel, whereas two-stage group testing algorithms and three-stage group testing algorithms are algorithms that consist of two stages and three stages, respectively, with each stage being a completely non-adaptive algorithm. In classical group testing, the potentially infected sets are all subsets of up to a certain number of elements of the given input set. For classical group testing, it is known that there exists a correspondence between classical superimposed codes and non-adaptive algorithms, and between two stage algorithms and selectors. Bounds on the number of tests for those algorithms are derived from the bounds on the dimensions of the corresponding combinatorial structures. Obviously, the upper bounds for the classical case apply also to our group testing model. In the present paper, we aim at improving on those upper bounds by leveraging on the characteristics of the particular hypergraph at hand. In order to cope with our version of the problem, we introduce new combinatorial structures that generalize the notions of classical selectors and superimposed codes.