A Novel Smoothing Framework for Interval-zero-order Takagi-Sugeno Fuzzy Neural Networks: Combining Interval Analysis with Zero-order TS Systems
Jian Li, Qiang Shao, Yuanquan Liu, Yiting Wang, Yan Liu · 2024
This study introduces a smoothing interval zero-order Takagi-Sugeno (SI-TS0) fuzzy neural network, a novel approach that combines interval numbers with the zero-order Takagi-Sugeno (TS0) fuzzy neural network. During experimentation, the interval TS0(I-TS0) fuzzy neural net-work led to significant oscillations in the gradient norm of the error function, primarily due to the non-differentiability of the absolute value function. To resolve this problem, a smoothing function was incorporated, successfully mitigating the oscillations and resulting in a more stable and accurate model. Extensive numerical simulations validate the SI-TS0 fuzzy neural network's effectiveness, demonstrating enhanced noise-handling capabilities and robustness, making it a valuable tool for processing fuzzy data in complex systems.