Belief Propagation With Optimized Pool Size for Non-Adaptive Group Testing: An Empirical Study

In this paper, an empirical study shows that positive tests containing multiple defectives are unlikely to provide effective messages in belief propagation (BP) for non-adaptive group testing. Thus, an objective function is proposed to measure the effectiveness of messages over edges, especially in the low-noise region. The maximization of the objective function allows us to optimize the pool size for BP. Simulation results show that the error performance of BP in the low-noise region is significantly improved by our pool size optimization.