In this paper, we derive mutual information based upper and lower bounds on the number of nonadaptive group tests required to identify a given number of "non-defective" items from a large population containing a small number of "defective" items. We show that a reduction in the number of tests is achievable compared to the approach of first identifying all the defective items and then picking the required number of non-defective items from the complement set. In the asymptotic regime with the population size N → ∞, to identify L non-defective items out of a population containing K defective items, when the tests are reliable, our results show that CsK 1−o(1) (Φ(α 0 , β 0 ) + o(1)) measurements are sufficient, where C s is a constant independent of N, K and L, and Φ(α 0 , β 0 ) is a bounded function of α 0 lim N →∞ L N −K and β 0 lim N →∞ K N −K . Further, in the nonadaptive group testing setup, we obtain rigorous upper and lower bounds on the number of tests under both dilution and additive noise models. Our results are derived using a general sparse signal model, by virtue of which, they are also applicable to other important sparse signal based applications such as compressive sensing.
Index TermsSparse signal models, nonadaptive group testing, inactive subset recovery. DRAFT to guaranteed detection of a small number of defective items. Two such properties were considered: disjunctness and separability [4]. 1 A probabilistic approach was adopted in [14]-[17], where random test matrix designs were considered, and upper and lower bounds on the number of tests required to satisfy the properties of disjunctness or separability with high probability were derived. In particular, [17] analyzed the performance of group testing under the so-called dilution noise. Another study [22] uses random test designs, and develops computationally efficient algorithms for identifying defective items from the noisy test outcomes by exploiting the connection with compressive sensing. A very recent work [25] uses novel information theoretic techniques, based on information density, to study the phase transitions for Bernoulli test matrix designs and measurement-optimal recovery algorithms. A general sparse signal model for studying group testing problems, that turns out to be very useful in dealing with noisy settings, was proposed and used in [18]- [21]. In this framework, the group testing problem was formulated as a detection problem and a one-to-one correspondence was established with a communication channel model. Using information theoretic arguments, mutual information based expressions (that are easily computable for a wide variety of noisy channels) for upper and lower bounds on the number of tests were obtained [21]. In the related field of compressive sensing, an active line of research has focused on the conditions under which reliable signal recovery from observations drawn from a linear sparse signal model is possible, for example, conditions on the number of measurements required and on isometry properties of the measur...