CAJ | 학술논문

BP神经网络（BP‐NN）因其自适应性、容错性和较强的泛化能力而得到广泛的研究及应用，但在实际应用中，却常常出现收敛速度慢、易陷入局部最优等问题。该文的新算法利用基因表达式编程（GEP）具有的良好全局搜索能力，对神经网络结构、权值及阈值进行优化；结合反向传播算法（BP）的局部搜索能力，有效提高了神经网络的性能；针对传统GEP设计神经网络会使网络结构失去层次性的问题，提出基于增加结构域染色体编码方法的GEP层次有序BP神经网络优化算法（GEPO‐NN），保证网络结构层次有序符合人脑分层处理模型：最后，通过仿真实验对比GEP和遗传算法（GA）对BP神经网络的优化性能。结果表明，GEPO‐NN有明显的性能提高。
BP신경망락（BP‐NN）인기자괄응성、용착성화교강적범화능력이득도엄범적연구급응용，단재실제응용중，각상상출현수렴속도만、역함입국부최우등문제。해문적신산법이용기인표체식편정（GEP）구유적량호전국수색능력，대신경망락결구、권치급역치진행우화；결합반향전파산법（BP）적국부수색능력，유효제고료신경망락적성능；침대전통GEP설계신경망락회사망락결구실거층차성적문제，제출기우증가결구역염색체편마방법적GEP층차유서BP신경망락우화산법（GEPO‐NN），보증망락결구층차유서부합인뇌분층처리모형：최후，통과방진실험대비GEP화유전산법（GA）대BP신경망락적우화성능。결과표명，GEPO‐NN유명현적성능제고。
BP neural network(BP‐NN)has been widely researched and applied because of its adapt‐ability ,fault tolerance and strong generalization capability .In practical application ,however ,it often appears with slow convergence speed and easily falls into local optimum problem .In this paper .by u‐sing the good global searching ability that gene expression programming (GEP) has ,neural network structure ,weights and thresholds are optimized ;Combined with the local search ability of back propa‐gation(BP)algorithm ,the performance of the neural network is effectively improved ;According to the problem that the traditional GEP designing can make the network structure suffer loss of order ,BP neural network orderly optimization algorithm based on GEP with structure domain (GEPO‐BP) is proposed to guarantee that the network structure should be orderly consistent with human hierarchical processing model .Finally ,by the experimental simulation of GEP and genetic algorithm (GA ) in the performance optimization of BP neural network ,the results show that the GEPO‐BP has obvious per‐formance improvement .