文献类型：专利

英文题名：Series-parallel mode identifying method for mechanical smell apparatus, involves forming classifier layer and function approximation model layer, and connecting classifier layer in series with function approximation model layer

作者：GAO D;SUN J;LIU F

机构：[1]UNIV EAST CHINA SCI & TECHNOLOGY

申请号：CN101101299-A

申请日：2007-06-25

公开日：2008-01-09

语种：英文

收录：DERWENT

摘要：NOVELTY - The method involves forming a classifier layer and function approximation model layer by parallel connection of multiple single outputting neural networks. The classifier layer is connected in series with the function approximation model layer to form an identifying model of a series-parallel mode based on the neural network. Training sub-class of the network is made up of samples denoting smell, and dummy of lopsided training sub-class is balanceable. Single output neural network of the classifier layer is corresponding to single-output neural network of the model layer. USE - Method for identifying a series-parallel mode of a mechanical smell apparatus. ADVANTAGE - The method permits the apparatus to realize real-time estimation of thousands of smell sorts and intensities at once. The method is simple, fast to learn, and the parameters determined by the method are less. The sort and function approximate precision is high in application of confirming mass smell sorts and intensities. DESCRIPTION OF DRAWING(S) - The drawing shows a schematic view illustrating a series-parallel mode identifying method.