基于核函数的SVM机与三层前向神经网络的关系_张铃.pdfVIP

25　 7 　　　　　　　　 Vol. 25 No. 7 2002 7 CHIN ESE J. COM PUTERS July 2002 　 SVM 张　铃 ( 　 230039) (　 100084) 　　 SV M, Vapnik SVM ; , ( ): () ,() , , () ,, ; , ,. 　,, : TP18 The Relationship Between Kernel Functions Based SVM and Three-Layer Feedforward Neural Networks ZHANG Ling (L aboratory of I ntel ligent Comp utin g S ignal Processing , I nstit ute of A rt if icial I ntell igence , A nhui University , H ef ei 230039) ( Sta te K ey L aboratory of I nt ell ig ent Technolog y Sy st ems, Tsingh ua Univer si ty , Bei ing 100084) Abstract　 In this paper, the equivalent between kernel functions based SV M (Vapnik) and the three layer feedforward neural network is demonstrated From the covering algorithms of neural. networks that author proposed, a kernel function existence theorem is proved. The theory show s that given a set of training samples, there must exist a corresponding function such that the im age of the training samples is linear separated in a high dimensional space under the mapping of the function. An algorithm of seeking the kernel functions is given. The computational complexi ty of the algorithm is polynomial growing with the sample size and the solution is the maximal margin one in the high dimensional space. Keywords　 kernal functions, SV M , three layer feedforward neural networks ; ( 3) 1　　 . 1995