Algorithmic implementation of deep learning layer assignment in edge computing based smart city environment

•This article proposes a Deep Learning layer assignment algorithm for edge computing environment (DLAEC) algorithm.•Conceptual breakthrough of proposed DLAEC is assigning DL layers with respect to resource capacity of individual nodes.•Simulated scenario is compared with multiple existing methods to confirm performance enhancement.•Results depicted a significant improvement in number of DL tasks and edge node resource utilization. Supporting deep learning is a challenge for Internet of Things hardware with limited computing capacity. Edge computing is a promising solution that supports such hardware as it solves transferring and processing bottlenecks. To allocate appropriate loads to utilize edge computing efficiently, we propose an edge computing solution for smart city environments that assigns some of the deep learning layers to edge nodes in order to support deep learning tasks on Internet of Things devices. The proposed deep learning layer assignment in edge computing algorithm determines the ideal number of deep learning layers to be assigned to each edge considering computing capacity and bandwidth of each edge separately. Simulation results of the proposed algorithm were compared with other existing methods such as Li's offline and online method, fixed assignment, and cloud-only method. The comparison showed that the proposed algorithm handles the most deep learning tasks while maximizing resource utilization of edges. [Display omitted]