Graph Reinforcement Learning-based CNN Inference Offloading in Dynamic Edge Computing

TL;DR

This paper introduces a graph reinforcement learning-based early-exit mechanism for CNN inference offloading in dynamic edge computing, significantly improving accuracy and throughput by effectively managing communication, computation, and accuracy trade-offs.

Contribution

It proposes a novel GRLE method that outperforms existing reinforcement learning approaches for CNN inference offloading in dynamic MEC environments.

Findings

GRLE achieves up to 3.41x higher accuracy than GRL.

GRLE outperforms DROOE with 1.45x higher accuracy.

Experimental results validate GRLE's effectiveness in dynamic scenarios.

Abstract

This paper studies the computational offloading of CNN inference in dynamic multi-access edge computing (MEC) networks. To address the uncertainties in communication time and Edge servers' available capacity, we use early-exit mechanism to terminate the computation earlier to meet the deadline of inference tasks. We design a reward function to trade off the communication, computation and inference accuracy, and formulate the offloading problem of CNN inference as a maximization problem with the goal of maximizing the average inference accuracy and throughput in long term. To solve the maximization problem, we propose a graph reinforcement learning-based early-exit mechanism (GRLE), which outperforms the state-of-the-art work, deep reinforcement learning-based online offloading (DROO) and its enhanced method, DROO with early-exit mechanism (DROOE), under different dynamic scenarios. The experimental results show that GRLE achieves the average accuracy up to 3.41x over graph reinforcement learning (GRL) and 1.45x over DROOE, which shows the advantages of GRLE for offloading decision-making in dynamic MEC.