摘要
针对标准BP神经网络在训练过程中,网络容易陷入局部极小点,使得进一步调整失去作用的问题,提出了一种有助于提高BP神经网络逼近精度的方法—基于模拟植物生长的学习算法,使BP网络在训练时能有效避开局部最小点,达到全局极小点,并保证了网络有较好的收敛速度。通过一个用Matlab编程的仿真实验表明了这一算法的有效性。
To overcome the lack of the standard BP algorithm that it easily fails into local minimum points in the training process, the paper brings forward a kind of training method of the BP network based on the plant growth simulation. The method put probability into the process of the network training to make the training to escape from the local minimum to reach the global minimum, and to render the network to speed up the convergence. One example is given to show the availability of the improved algorithm by using Matlab. The simulation result shows that the improved algorithm is available.
出处
《大连大学学报》
2008年第3期37-41,共5页
Journal of Dalian University
基金
国家自然科学基金(70371051)
关键词
BP算法
神经网络
模拟植物生长
BP algorithm
neural network
plant growth simulation algorithm