Towards Neural Architecture Search for Transfer Learning in 6G Networks
Adam Orucu, Farnaz Moradi, Masoumeh Ebrahimi, Andreas Johnsson
TL;DR
The paper addresses the challenge of automating neural architecture search (NAS) for transfer learning (TL) in AI-native 6G networks. It surveys how NAS and TL can be combined to adapt models under varying tasks, data regimes, and resource constraints, and highlights three key research directions: joint NAS-TL, multi-objective optimization, and handling tabular networking data. Core contributions include a structured background on TL and NAS, a synthesis of existing NAS-TL collaborations, and a delineation of open challenges with concrete research directions and ongoing work centered on automatic model lifecycle management for 6G networks. The work aims to enable scalable, efficient, and adaptable ML deployments across devices, edge, and cloud resources in future networks. Overall, it provides a roadmap for advancing NAS-based TL approaches to meet the stringent, dynamic requirements of 6G infrastructure and services.
Abstract
The future 6G network is envisioned to be AI-native, and as such, ML models will be pervasive in support of optimizing performance, reducing energy consumption, and in coping with increasing complexity and heterogeneity. A key challenge is automating the process of finding optimal model architectures satisfying stringent requirements stemming from varying tasks, dynamicity and available resources in the infrastructure and deployment positions. In this paper, we describe and review the state-of-the-art in Neural Architecture Search and Transfer Learning and their applicability in networking. Further, we identify open research challenges and set directions with a specific focus on three main requirements with elements unique to the future network, namely combining NAS and TL, multi-objective search, and tabular data. Finally, we outline and discuss both near-term and long-term work ahead.