Reproduce “Edge Intelligence: Architectures, Challenges, and Applications” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Dianlei Xu, Tong Li, Yong Li, Xiang Su, Sasu Tarkoma, Tao Jiang, Jon Crowcroft, Pan Hui
This survey offers a comprehensive, micro-view analysis of Edge Intelligence, defining its four fundamental components: edge caching, training, inference, and offloading. It systematically classifies existing solutions, techniques, and applications while also outlining critical open challenges and future research directions.
Key insights
- Realizing effective edge intelligence necessitates addressing interdependent challenges across data management (caching), distributed model development (training), efficient model deployment (inference), and flexible resource utilization (offloading).
- Practical edge AI solutions extensively leverage techniques such as model compression, federated learning for privacy-preserving distributed training, and dynamic offloading strategies to balance computational loads across devices, edge servers, and the cloud.
- Major ongoing research challenges include managing data scarcity and consistency at the edge, enhancing model adaptability to dynamic environments, ensuring robust privacy and security, and designing effective incentive mechanisms for device participation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.