LENS: Layer Distribution Enabled Neural Architecture Search in Edge-Cloud Hierarchies
Hosted in Virtual Platform
Design of Cyber-physical Systems and IoT
DescriptionEdge-Cloud hierarchical systems employing intelligence through Deep Neural Networks (DNNs) endure the dilemma of workload distribution within them. Previous solutions proposed distributing workloads at runtime according to the state of surrounding conditions, like the wireless connection. However, such conditions are usually overlooked at design time. This paper addresses that for DNN architectural design by presenting a novel methodology that administers multi-objective Neural Architecture Search over two-tiered systems, where the performance-related objectives are refashioned to consider the expected wireless communication parameters. We demonstrate that our methodology identifies more efficient deployment options compared to traditional methodologies for edge-cloud hierarchies.