A Scalable Federated Modular Neural Network Architecture for Edge-Enabled Beyond-5G Complex Systems

Artificial intelligence (AI) continues to be transformative across various domains, integrating learning paradigms like machine learning, modular learning, and federated learning. In conjunction with emerging technologies such as the Internet of Things, Industry 4.0, blockchain-enabled Web 3.0, and 5G, AI has fueled numerous real-world applications. However, modern industrial systems face complex challenges with heterogeneous requirements that exceed the capabilities of conventional neural networks. To address this, we propose an AI-based modular architecture reinforced by blockchain and 5G network slicing services. Our approach utilizes federated modular strategies driven by intelligent storage systems to decompose problems into subsystems using AI and other techniques. This system automates processes, reducing errors and improving network efficiency, scalability, and data generalization. We exemplify our methodology through a health-care use case, demonstrating its efficacy.