Approximation of regular fuzzy neural networks based on Sugeno integral

We study approximation of fuzzy neural networks on regular subadditive fuzzy measure spaces. We shall show that four-layer regular fuzzy neural networks can approximate any fuzzy-valued measurable function for fuzzy integral (Sugeno integral) and the Shilkret integral norm, respectively. Since the continuity of fuzzy measures is not required, the previous results obtained by other researchers are generalized.