Distributed Deep Neural Networks with System Cost Minimization in Fog Networks
Syed Danial Ali Shah, Hong Ping Zhao, Hoon Kim · 2018
This paper proposes an efficient network resource allocation in neural networks over hybrid computing hierarchies, consisting of the cloud, the fog and end devices (hybrid-fog networks, H-FogN). Firstly, we introduce network architecture and network components for intelligence delivery in which network resources for computing, communication, and storage are embedded. We formulate the problem of resource allocation and computation offloading for load reduction in the cloud center for fog networks. The goal is to minimize the maximum cost associated with the communication for all the users while meeting the requirements of maximum delay and bandwidth consumption constraints.