Efficient and Accurate Group Testing via Belief Propagation: An Empirical Study

The group testing problem asks for efficient pooling schemes and inference algorithms that allow to screen moderately large numbers of samples for rare infections. The goal is to accurately identify the infected individuals while minimizing the number of tests. We propose the novel adaptive pooling scheme adaptive Belief Propagation (ABP) that acknowledges practical limitations such as limited pooling sizes and noisy tests that may give imperfect answers. We demonstrate that the accuracy of ABP surpasses that of individual testing despite using few overall tests. The new design comes with Belief Propagation as an efficient inference algorithm. While the development of ABP is guided by mathematical analyses and asymptotic insights, we conduct an experimental study to obtain results on practical population sizes.