An algorithm for global optimization of distributed multiple-sensor detection systems using Neyman-Pearson strategy

In 1986, Srinivasan derived a set of strongly coupled conditions defining solutions to a Lagrange multiplier optimization problem which is sometimes equivalent to the Neyman-Pearson distributed hypothesis testing problem. These conditions have been of limited use because they are difficult to solve. An algorithm for solution of Srinivasan's conditions is proposed. The algorithm scans all possible solutions and picks the optimal one in the Neyman-Pearson sense. It was applied to examples of distributed-detection systems with exponentially distributed observations and was found to be simple, accurate and fast.