General Regression Neuro–Fuzzy Network for Identification of Nonstationary Plants

In this work, we propose a General Regression Neuro-Fuzzy Network that combines the characteristics of traditional General Regression Neural Networks and Adaptive Network-based Fuzzy Inference Systems. This innovative approach addresses the challenges of identifying nonstationary plants, particularly in scenarios where data must be processed in real-time alongside the plant's operation. By employing a one-pass learning algorithm, the proposed network optimally adjusts its parameters to effectively handle the complexities of real-time data feeds. We explore the limitations of conventional multilayer perceptrons in situations involving nonstationary plants and demonstrate the advantages of using Radial Basis Function Networks (RBFN) as an alternative. Additionally, the implementation of Normalized Radial Basis Function Networks (NRBFN) mitigates common issues such as the curse of dimensionality and inactive neuron regions through space partition of unity techniques. Ultimately, our General Regression Neuro-Fuzzy Network enhances the effectiveness of express-diagnostic devices in assessing plant health and environmental factors, thereby contributing significantly to the field of biosensor technology.