Keynote Speaker

Prof. Dianhui Wang, IEEE Senior Member, Qingdao University of Science and Technology, China
Bio: Prof. Dianhui Wang is a National Distinguished Expert, Professor, Doctoral Supervisor, and IEEE Senior Member at the School of Data Science, Qingdao University of Science and Technology (QUST), China, and Skolkovo Institute of Science and Technology (Skoltech), Russia. He received his Ph.D. in Industrial Automation from Northeastern University, China, in 1995. His research interests include stochastic configuration learning, lightweight artificial intelligence for industrial applications, complex industrial big data modeling, edge intelligence, and prediction and early warning of rare critical events in industrial systems. He is the founder of Stochastic Configuration Networks (SCNs) and Stochastic Configuration Machines (SCMs), with extensive research contributions to intelligent modeling and predictive analytics for complex industrial systems. He is also the Editor-in-Chief of the international journal Industrial Artificial Intelligence (Springer Nature) and has served as an Associate Editor for several IEEE and other internationally renowned academic journals.
Speech Title: Stochastic Configuration Networks: Lightweight Learning Paradigms for Edge Intelligence and Industrial Service Computing
Abstract: With the rapid proliferation of the Internet of Things (IoT) and cyber-physical systems, migrating artificial intelligence from centralized cloud data centers to resource-constrained edge devices has become a critical imperative. However, mainstream deep learning architectures often suffer from excessive computational overhead, slow training convergence, and poor adaptability to dynamic streaming data, which severely hinder their deployment at the network edge.
In this talk, we introduce Stochastic Configuration Networks (SCNs)—a class of randomized learner models with universal approximation capabilities—as an efficient foundation for edge intelligence. Unlike traditional randomized neural networks, SCNs incrementally construct hidden nodes under a supervisory mechanism that adaptively determines the scope of random parameters, guaranteeing fast convergence and compact network structures. We will explore how SCNs and their variants, including Stochastic Configuration Machines (SCMs), enable ultra-fast on-device training, real-time incremental learning, and low-latency inference on edge hardware. Furthermore, practical industrial applications such as complex system modeling will be demonstrated, illustrating how lightweight randomized learning empowers reliable and responsive edge service computing.