摘要
对普通并联神经元的缺陷进行了分析,提出了一种广义的并联抑制神经元,构造了基于并联抑制神经元的前向神经网络结构,并给出了相应的学习算法。通过对几个模式分类问题的基准问题的测试,将提出的方法与SIANN、BP神经网络进行了比较,验证了提出的网络结构和学习算法的有效性。实验结果表明:单个的GSIN和简单的GSINN可以取得比SIANN和BP网络都好的分类效果。
A Generalized Shunting Inhibition Neuron （GSIN） model is proposed by analyzing shortcomings of the normal shunting neuron model. A new feed forward neural network architecture based on GSIN, naming Generalized Shunting Inhibition Neural Network （GSINN）, and its learning algorithm are then introduced. Finally, the GSINN is applied to several benchmark classification problems, and their performance is compared with the performances of Shunting Inhibitory Artificial Neural Network （SIANN） and BP networks, and the effectiveness of the proposed network structure and learning algorithm is verified. Experimental results show that a single GSIN and simple GSINN can outperform both the SIANN and Back Propagation （BP） network.
出处
《电子科技大学学报》
EI
CAS
CSCD
北大核心
2006年第3期399-402,共4页
Journal of University of Electronic Science and Technology of China
基金
国家自然科学基金资助项目(60135010)
关键词
BP网络
模式分类
并联抑制
学习算法
BP network
pattern classification
shunting inhibition
learning algorithm
