Paper in proceedings, 2012

The classical group testing problem asks to determine at most d defective elements in a set of n elements, by queries to subsets that return Yes if the subset contains some defective, and No if the subset is free of defectives.
By the entropy lower bound, d log n tests are needed at least. We devise group testing strategies that combine two features: They achieve this optimal query bound
asymptotically, with a factor 1+o(1) as n grows, and they work in a fixed number of stages of parallel queries. Our strategies are randomized and have a controlled failure probability, i.e., constant but arbitrarily small. We
consider different settings (known or unknown d, probably correct or verified outcome), and we aim at the smallest possible number of stages. In particular, 2 stages are sufficient if d grows slowly enough with n, and 4 stages are sufficient if d=o(n).

bioinformatics

randomized algorithm

parallel queries

group testing

algorithmic learning theory

Chalmers, Computer Science and Engineering (Chalmers), Computing Science (Chalmers)

Chalmers, Computer Science and Engineering (Chalmers), Computing Science (Chalmers)

03029743 (ISSN) 16113349 (eISSN)

Vol. 7147 214-225Basic sciences

Life Science Engineering (2010-2018)

Computer Science

10.1007/978-3-642-27660-6_18

978-3-642-27659-0