LAUSR

Abstract In this paper, we present an adaptive neuro-complex-fuzzy-inferential modeling mechanism (ANCFIMM) for generating higher-order Takagi-Sugeno-Kang (TSK) models, and exploit its split complex-valued gradient descent algorithm (SCGDA). ANCFIMM has five layers, among which the Gaussian membership layer casts the M-dimensional complex-valued imported features to a K-dimensional real number field. Hence, we use the splitting strategy to acquire the gradients of the real-valued function in deriving the learning algorithm of ANCFIMM. The monotonicity and deterministic convergence of SCGDA are demonstrated and proved in detail. Experiments are also implemented to manifest the properties of ANCFIMM on a series of data sets, which effectively validate the theoretical observations.