分式线性神经网络及其非线性逼近能力研究杨国为.pdfVIP

30 　2 　　　　　　　　 Vol. 30 No. 2 2007 2 CHINESE JOURNAL OF COMPUTERS Feb. 2007 　 1), 2) 2) 3) 杨国为 　王守觉 　闫庆旭 1)(　　266071) 2)(　　100083) 3)(　　100083) m n 　　, R R m n , , R R , BP ;. . . 　;;; TP183 Research of Fractional Linear Neural Network and Its Ability for Nonlinear Approach 1), 2) 2) 3) ANG Guo-Wei 　WANG Shou-Jue 　 AN Qing-Xu 1)(Institute of Pattern Recogn ition and In telligent Sy stem, Qing dao University , Qingdao　266071) 2)(Laboratory of Artif icial N eural N etwork, Institute of S emiconductors, C inese Academy of S ciences, B eij ing 　100083) 3)(S c ool of Inf ormation Eng ineering, C ina University of Geosciences, Beij ing 　100083) Abstract 　 This paper presents a fractional linear neural network with simple structure, and proves that the fractional linear neural network can limitlessly approach any continuous mapping m n from limited close subset of R to R . T he neural network also can limitlessly approach any con- m n tinuous mapping, which has limit at infinite place, from limitless close subset of R to R . It ex- tends the nonlinear approach ability of traditional BP neural network. Moreover the back propa- gation algorithm for a fractional linear n