Workflow from Scientific Research

The illustration aims to summarize the large variety of operations that ShELL requires with respect to conventional ML approaches. The four main categories of operations are shown. Specific operations for each category (grey boxes), and their hardware allocation (0 to 5) are also shown. The blue shades highlight neural-network-specific computation that mostly characterizes ShELL on edge devices. Such a computation for LL and sharing benefits from special hardware such as GPU, field-programmable gate arrays (FPGAs) and application-specific integrated circuits (ASICs) to facilitate deployment of LL algorithm to SWaP-constrained devices.

DOI