摘要
采用Voronoi向量对SOM网络算法进行了改进 ,提高了学习收敛速度。通过提取数据的统计特征 ,建立了可靠性分布模式自动识别样本。提出的智能自动识别模型分两层 ,在SOM网络层对概率分布模式进行自动聚类 ,在支持向量机层对各聚类组进行分类学习和识别 ,获得识别模型的双层记忆权值。最后采用模型对常用可靠性分布模式进行了自动识别研究。测试结果表明 ,建立的可靠性分布模式自动识别模型是可行。
An intelligent identification combined structure model is proposed using self-organizing map (SOM) and support vector machines (SVM). This model can improve the self-organizing map algorithm using Voronoi vector to reduce space occupation and improve convergence, and develop probability intelligent identification training samples set. Due to the complexity of the summary statistics, the authors select kurtosis, skewness, quantile and cumulative probability as parameters for data distributions identification training sets in experience. The combined structure model is divided into two layers. In the first SOM layer, different reliability distributions training sets are clustered into groups using SOM. In the second SVM layer, the clusters are learned and classified respectively in each group using novel multi-class support vector machines. Random data time series of 23 types of probability distributions are identified through the test in the trained model. The results indicate that the identification rates by the intelligent model are higher as compared to those achieved by BP neural networks and probability networks models.
出处
《航空学报》
EI
CAS
CSCD
北大核心
2003年第3期207-211,共5页
Acta Aeronautica et Astronautica Sinica
基金
国防预研资助基金 (项目编号 :98J19.3.2 .JB32 0 1)
空军重点型号工程课题资助
关键词
神经网络
支持向量机
机器学习
可靠性
概率分布
模式识别
Computer simulation
Neural networks
Pattern recognition
Probability distributions
Safety testing
Self organizing maps
Self organizing storage
