摘要: 提出了一种基于增长法的神经网络结构优化算法。在函数逼近的BP神经网络中引入一种改进的BP算法(LMBP算法),通过二次误差下降与梯度下降,利用误差变化规律分析网络结构的优化程度,自适应地增加隐层神经元或网络层次,从而得到一个合适的网络结构。进行了仿真实验及该算法与RAN算法用于逼近函数的对比实验,实验结果表明了该算法的有效性。
关键词:
神经网络,
结构优化,
LMBP算法,
函数逼近,
RAN算法
Abstract: An algorithm to optimize artificial neural networks structure based on constructive method is presented. And an improved BP algorithm, LMBP algorithm, is introduced about simplest BP neural network for function approximation. By using the rules of error changing based on quadratic error and gradient reducing, this paper analyzes the optimization of the network structure. By adding hidden neurons or network layers adaptively, a proper structure of the network is got. Simulation experiments are provided to compare the approach with RAN algorithm for solving function approximation. The results show the effectiveness of the algorithm.
Key words:
neural networks,
structure optimization,
LMBP algorithm,
function approximation,
RAN algorithm
中图分类号:
杨 英;唐 平;王越超;丘衍航. 基于LMBP改进算法的神经网络结构优化[J]. 计算机工程, 2008, 34(1): 215-217.
YANG Ying; TANG Ping; WANG Yue-chao; QIU Yan-hang. Neural Network Structure Optimization Based on Improved LMBP Algorithm[J]. Computer Engineering, 2008, 34(1): 215-217.