摘要：AbstractFractional-order extreme learning machine is presented for a regression problem in this paper. The objective optimization function is designed as a fractional-order one, which can obtain better optimization results in training and testing errors. Moveover, Lévy flight which is a class of fractional-order continuous random walk owing a heavy-tailed distribution, is used to search the related weight and threshold parameters. The connection between the order of Lévy flight and the error result is discussed. Some given examples are provided to verify the better performance of the proposed methods.
