Date: 04 Jun 2002

Improved Algorithms for Chemical Threshold Testing Problems

* Final gross prices may vary according to local VAT.

Get Access

Abstract

We consider a generalization of the classical group testing problem. Let us be given a sample contaminated with a chemical substance. We want to estimate the unknown concentration c of this substance in the sample. There is a threshold indicator which can detect whether the concentration is at least a known threshold. We consider either the case when the threshold indicator does not affect the tested units and the more difficult case when the threshold indicator destroys the tested units. For both cases, we present a family of efficient algorithms each of which achieves a good approximation of c using a small number of tests and of auxiliary resources. Each member of the family provides a different tradeoff between the number of tests and the use of other resources involved by the algorithm. Previously known algorithms for this problem use more tests than most of our algorithms do. For the case when the indicator destroys the tested units, we also describe a family of efficient algorithms which estimates c using only a constant number of tubes.