Improving the Reliability of Pooled Testing with Combinatorial Decoding and Compressed Sensing

The problem of detecting few viral infections in a possibly large group with as few as possible tests can be modeled as either a group testing (GT) or a compressed sensing (CS) problem. CS approaches also allow to recover the viral load, but the underlying noise models are not common in CS and not well understood. Therefore, we study hybrid approaches that combine methods from CS and GT on various noise models. We compare the performance of such approaches with classical decoders from CS and GT. Our results show that combined strategies can improve the error rates and provide viral load estimation.