An Efficient Learning Algorithm for Large-Scale Feedforward Neural Networks
One impediment to use of neural networks in modelling the behavior of complex systems is the excessive time required for supervised learning in larger multilayer feedforward networks. The use of nonlinear optimization techniques to perform neural network training offers a means of reducing that computing time. In this paper, a neural network algorithm based on modified quasi-Newton method is introduced, aiming at enhancing the neural networks ability to solve the modeling problem of large-scale. Compared with quasi-Newton method, this algorithm needs less memory and convergence is almost the same. The proposed methodology is applied to modeling of industrial product quality with 32-input. Simulation results show the effectiveness of the approach.
Huanqin Li Xiaohua Li
Faculty of Science Xian Jiaotong University Xian 710049, China Signal & Information Processing Lab Bejing University of Technology Beijing,100022,China
国际会议
南宁
英文
2007-07-20（万方平台首次上网日期，不代表论文的发表时间）