ConGCNet: Convex geometric constructive neural network for Industrial Internet of Things

• DOI: 10.1016/j.jai.2024.07.004
• Publication Date: 2024-07-20T16:51:09Z
• Authors: Jing Nan; Wei Dai; Chau Yuen; Jinliang Ding
• Keywords: Constructive; Benchmark (surveying)

The intersection of the Industrial Internet Things (IIoT) and artificial intelligence (AI) has garnered ever-increasing attention research interest. Nevertheless, dilemma between strict resource-constrained nature IIoT devices extensive resource demands AI not yet been fully addressed with a comprehensive solution. Taking advantage lightweight constructive neural network (LightGCNet) in developing fast learner models for IIoT, convex geometric low-complexity control strategy, namely, ConGCNet is proposed this article via optimization matrix theory, which enhances convergence rate reduces computational consumption comparison LightGCNet. First, strategy to reduce during hidden parameters training process. Second novel output weights evaluated method based on guarantee rate. Finally, universal approximation property proved by method. Simulation results, encompassing four benchmark datasets real ore grinding process, demonstrate that effectively modelling process improves model's