This paper proposes a standpoint of fusion development of support vector machines (SVM) and neural networks (NN), and argues that SVM and NN are equivalent in function but different in implementation algorithm. In term of cognitive model, it discusses the guidance effect of NN cognitive model on SVM counterpart development, and simultaneously investigates the concepts and development methods of SVM cognitive model in detail. Finally, it proposes a class of NN algorithm development from the angle of core idea of SVM algorithm.