An efficient algorithm for combinatorial group testing

Abstract

In the (d,n) group testing problem n items have to be identified as either good or defective and the number of defective items is known to be d. A test on an arbitrary group (subset) of items reveals either that all items in the group are good or that at least one of the items is defective, but not how many or which items are defective. We present a new algorithm which in the worst case needs less than tests more than the information lower bound for. For, the difference decreases to less than tests. For d ≥ 10, this is a considerable improvement over the d - 1 additional tests given for the best previously known algorithm by Hwang. We conjecture that the behaviour for large n and d of the difference is optimal for. This implies that the tests per defective given in the bound above are the best possible. © Springer-Verlag Berlin Heidelberg 2013.