Efficient Edge Computing: Harnessing Compact Machine Learning Models for Workload Optimization

Mohammad Reza Pour-Hosseini, Mahdi Abbasi, Bahman Javadi, Seifeddine Benelghali, Hassan Haghighi, Ehsan Mohammadi Pasand, Parham Moradi

Abstract: This paper proposes an innovative lightweight workload distribution model, TinyXCS, specifically designed for traditional edge devices, emphasizing efficiency and compactness. TinyXCS, an online model based on XCS learning classifier systems, demonstrates high levels of performance in reducing delays and energy consumption. Engineered to operate efficiently on conventional edge devices, The model offers a promising solution for optimizing workload distribution while considering memory constraints. Experimental results validate effectiveness of this model, representing a significant advancement in the quest for streamlined edge computing solutions.