Energy-Efficient DNN Partitioning and Offloading for Task Completion Rate Maximization in Multiuser Edge Intelligence

Deep Neural Network (DNN) has become an essential technology for edge intelligence. Due to significant resource and energy requirements large-scale DNNs’ inference, executing them directly on energy-constrained Internet of Things (IoT) devices is impractical. DNN partitioning provides a feasible solution this problem by offloading some layers execute the server. However, resources servers are also typically limited. An resource-constrained optimization generated in such realistic environment. Motivated this, we investigate multiuser environment, which considered intractable Mixed-Integer Nonlinear Problem (MINLP). We decompose into two subproblems propose Energy-Efficient Partitioning Offloading (EEDPO) strategy solve it polynomial time based minimum cut/maximum flow theorem dynamic programming. Finally, test impact constraint, type, device number performance EEDPO. Simulation results models demonstrate that proposed can significantly improve inference task completion rate compared other methods.