Neural Architecture Search for Efficient Edge Computing: A Multi-Objective Optimization Approach

Posted: Oct 28, 2025

• Wei Zhang
• Maria Rodriguez
• Kenji Tanaka
• Fatima Al-Mansoori

Abstract

This paper presents a novel neural architecture search (NAS) framework specifically designed for edge computing environments where computational resources and power consumption are critical constraints. We propose a multi-objective optimization approach that simultaneously minimizes model size, inference latency, and computational complexity while maintaining competitive accuracy on image classification tasks. Our methodology employs a modified evolutionary algorithm with adaptive mutation rates and incorporates hardware-aware performance metrics directly into the search process. Experimental results on CIFAR-10 and ImageNet datasets demonstrate that our approach discovers neural architectures that achieve 94.7% accuracy on CIFAR-10 with only 2.1M parameters and 15ms inference time on Raspberry Pi 4 hardware, representing a 3.2× improvement in efficiency compared to manually designed architectures. The proposed framework provides a systematic approach to developing efficient deep learning models for resource-constrained environments.