Power scheduling for distributed multiple-hypothesis detection by task-specific information

We introduce a new information theoretic power allocation scheme applicable to distributed multiple-hypothesis detection systems communicating over slow fading channels. In earlier work, it was demonstrated that performance could be improved by adjusting transmit power to maximize the J-divergence measure of a binary detection system and the J-divergence method is extended for a distributed multiple-hypothesis detection system by defining pairwise sums of the J-divergences. However, the pairwise sum measure does not provide a tight bound. Basically, the more hypotheses we adopt, the less efficient the optimization is. Thus, we derive a more efficient classification-oriented information measure for power optimization of distributed multiple-hypothesis system by introducing a virtual decider variable. The virtual decider variable is directly related with classification task. Various numerical results are also shown to compare the performances.