C5 · Conference Paper Published

RedgeX: Meta-learning based Optimal Analytical Model for Programmable Edge Intelligence

Mehbub Alam, Nurzaman Ahmed, Rakesh Matam and Ferdous Ahmed Barbhuiya

IEEE Wireless Communications and Networking Conference (WCNC), pp. 1-6, 2024

Publisher Page DOI View Presentation Cite Paper

Abstract

In this paper, we propose RedgeX, a meta-learning based approach for generating analytical models in a distributed edge intelligence network. The approach involves training a meta-learning model on a large dataset of edge device information and performance metrics to predict the optimal analytical model for a given task and available resources. An edge controller, which has the status of all the edge devices, can then deploy the optimal model to the most suitable edge devices based on their available resources. The RedgeX improves the efficiency and effectiveness of edge intelligence systems by dynamically generating analytical models based on the specific requirements of each task and the available resources in the edge devices. The performance evaluation of the proposed scheme shows better utilization of resources, improved performance, and reduced latency in edge intelligence systems.

Research Highlights

  1. We introduce a meta-learning based analytical model generation technique for edge devices. The approach makes informed decisions by considering the states and resource availability of edge devices, optimizes the selection of code snippets, generates an analytical module from them, and remotely deploys it to the target edge device.
  2. We introduce a global-status driven model for edge computing and intelligence, employing an edge controller that regularly gathers status updates from edge devices to maintain a current and comprehensive overview of resource availability across the network's nodes.

Presentation

A 5-slide visual summary of this paper. Use the arrows, the dots or the left/right arrow keys to move between slides.

Slide 1 of 5 from the visual summary of “RedgeX: Meta-learning based Optimal Analytical Model for Programmable Edge Intelligence”
Slide 2 of 5 from the visual summary of “RedgeX: Meta-learning based Optimal Analytical Model for Programmable Edge Intelligence”
Slide 3 of 5 from the visual summary of “RedgeX: Meta-learning based Optimal Analytical Model for Programmable Edge Intelligence”
Slide 4 of 5 from the visual summary of “RedgeX: Meta-learning based Optimal Analytical Model for Programmable Edge Intelligence”
Slide 5 of 5 from the visual summary of “RedgeX: Meta-learning based Optimal Analytical Model for Programmable Edge Intelligence”
5 slides
Download slide deck (PDF)