PCR testing is an invaluable diagnostic tool that has most recently seen widespread use during the COVID-19 pandemic. A recent work by Wang, Gabrys and Vardy proposed tropical codes as a model for group PCR testing. For a known but arbitrary number of infected persons, a sufficient condition on the underlying block design of a zero-error tropical code, called double disjunction, is proposed. Despite this, the parameters for which the construction of doubly disjunct block designs is known to exist are very limited. In this paper, we define probabilistic tropical codes and consider random block designs that are doubly disjunct with high probability. We also provide a deterministic construction for a doubly disjunct block design given a disjunct block design. We show that for certain choices of parameters, our probabilistic construction has vanishing error. Our constructions, combined with existing methods, give us three different ways to construct tropical codes. We compare the number of tests required by each, and bounds on the error.
翻译:暂无翻译