Optimizing Edge AI for Real-Time Decision- Making: A Hybrid Approach Using Model Compression and Federated Learning

Abstract

Edge AI is transforming real-time decision-making by enabling intelligent data processing directly on edge devices. However, its full potential is limited by challenges like computational constraints, latency, and energy efficiency. This research presents a hybrid approach that combines model compression and federated learning to enhance Edge AI performance. Techniques such as quantization and pruning are applied to minimize computational load while preserving accuracy. Federated learning enables secure, privacy-focused collaborative training without sharing raw data, improving both security and efficiency. The proposed framework is tested on benchmark datasets, showing enhancements in processing speed, energy efficiency, and inference accuracy. Experimental findings highlight the balance between compression ratios, model accuracy, and training efficiency, offering insights into optimal implementation. This study contributes to the advancement of Edge AI in resource-limited environments, including autonomous systems, healthcare, and IoT applications.