Randomized fusion rules can be optimal in distributed Neyman-Pearson detectors
Yong In Han, Taejon Kim · IEEE Transactions on Information Theory · 1997
We show that randomized fusion rules can be locally optimal in distributed detection systems under the Neyman-Pearson criterion. This result is contrary to common belief. We first formulate conditions for a randomized fusion rule to be locally optimal. Then, we present distribution functions of local observations that satisfy these conditions.