Neural Architecture Search for Efficient Convolutional Networks: A Multi-Objective Optimization Approach

Posted: Apr 15, 2024

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

Abstract

This paper presents a novel neural architecture search (NAS) framework that optimizes convolutional neural networks for both accuracy and computational efficiency. Our approach employs a multi-objective evolutionary algorithm to explore the architecture space, balancing model performance with resource constraints. We introduce a hierarchical search space representation that enables efficient exploration of network depth, width, and connectivity patterns. Experimental results on CIFAR-10 and ImageNet datasets demonstrate that our method discovers architectures that achieve competitive accuracy while reducing computational requirements by up to 45% compared to hand-designed networks. The proposed framework provides a systematic approach to automated neural network design, addressing the growing need for efficient deep learning models in resource-constrained environments.