Graph-based Deep Learning for Communication Networks: A Survey

Abstract: Communication networks are important infrastructures in contemporary society. There are still many challenges that are not fully solved and new solutions are proposed continuously in this active research area. In recent years, to model the network topology, graph-based deep learning has achieved the state-of-the-art performance in a series of problems in communication networks. In this survey, we review the rapidly growing body of research using different graph-based deep learning models, e.g. graph convolutional and graph attention networks, in various problems from different types of communication networks, e.g. wireless networks, wired networks, and software defined networks. We also present a well-organized list of the problem and solution for each study and identify future research directions. To the best of our knowledge, this paper is the first survey that focuses on the application of graph-based deep learning methods in communication networks involving both wired and wireless scenarios. To track the follow-up research, a public GitHub repository is created, where the relevant papers will be updated continuously.

• The paper presents a comprehensive survey of graph-based deep learning applications in communication networks from 2016 to 2021.
• The study demonstrates that GNNs effectively model spatial topologies, enabling robust decision-making in wired, wireless, and SDNs.
• The review synthesizes 81 studies and outlines future research directions to integrate advanced AI techniques for network optimization.

Graph-based Deep Learning for Communication Networks: A Survey

The paper "Graph-based Deep Learning for Communication Networks: A Survey" provides a comprehensive analysis of the emerging role of graph-based deep learning models in addressing various challenges within communication networks. The study acknowledges the crucial function of communication networks as societal lifelines but also highlights the dynamic challenges that persist and evolve within this domain. The authors focus on graph-based deep learning methods, notably Graph Neural Networks (GNNs), which stand out for their efficacy in modeling network topology and solving complex problems characteristic of both wired and wireless networks.

Key Insights and Contributions

1. Scope and Coverage: The survey extensively covers studies from 2016 to 2021, examining a vast array of graph-based deep learning models like graph convolutional networks (GCNs) and graph attention networks (GATs). It categorizes applications across typical network types, including wireless, wired, and software-defined networks (SDNs), thereby providing a panoramic view of the field.
2. Highlight on GNNs: GNNs represent the cornerstone of this work, distinguished by their ability to model spatial information inherent in network topologies and adapt to dynamic network changes. The paper delineates how GNNs facilitate effective decision-making and problem-solving in scenarios that traditional deep learning models struggle to address due to network complexity and topology variability.
3. Applications and Numerical Results: The survey identifies successful GNN applications across an array of network challenges:
   • Wireless Networks: GNNs have been instrumental in optimizing resource allocation, power control, and traffic prediction within cellular and ad hoc networks.
   • Wired Networks: The potential of GNNs is realized in network modeling, routing optimization, and security enhancement, proving particularly effective in complex environments like data centers and optical networks.
   • Software-Defined Networks: The decoupled architecture of SDNs combined with GNN capabilities has enabled enhanced network design and management solutions, including routing optimization and virtual network function placement.
4. Comprehensive Synthesis of Literature: The authors provide a well-organized summary of 81 studies, bringing clarity to the various problems addressed and solutions proposed, while also maintaining a repository for continuous literature updates via a public GitHub repository.

Implications and Future Directions

The implications of this survey are substantial for both theoretical advancements and practical implementations of communication networks. The paper underscores the transformative potential of integrating graph-based methodologies, especially within contexts that demand robust and scalable solutions due to dynamic topology changes or increased scale.

Looking forward, the paper suggests potential research directions, including the integration of GNNs with other AI techniques such as generative adversarial networks (GANs) and automated machine learning (AutoML), scaling applications to larger network topologies, and further exploration of novel GNN variants to solve yet-unaddressed challenges within future communication architectures.

In summary, this survey positions graph-based deep learning at the forefront of future-proofing communication networks, encouraging a paradigm shift towards more intelligent, adaptable, and efficient network management and operation strategies using state-of-the-art deep learning models. As the field evolves, further research leveraging the insights from this survey could catalyze advancements across a spectrum of networking technologies and applications.