Title: Self-organized mapping network based document clustering method
Application Number: 200610009761 Application Date: 2006.03.02
Publication Number: 1808474 Publication Date: 2006.07.26
Approval Pub. Date: Granted Pub. Date:
International Classifi-cation: G06N3/08,G06F15/18,G06F17/30
Applicant(s) Name: Harbin Institute of Technology Address: 150001
Inventor(s) Name: Liu Yuanchao, Guan Yi, Xu Zhiming, Liu Bingquan
Attorney & Agent: mao yonglin
Abstract:
     The file clustering method based on self-organization mapping network comprises: finding out all the selected files; initiating the output layer of the network as ring structure to divide it equally at least into two parts with every sector as a neuron; calculating R2 clustering criterion factor of current output layer; deciding whether the said factor more than the threshold mu, if yes, stopping the training and classifying the file according to current self-organization mapping network; or else, finding out the neuron with maximal RSS in class to interpolate a new neuron, then training all neurons of current output layer. This invention can self-adapt the input file data and fixed structure and other incidental problems.
Time: 6
<- Previous Patent:Fuzzy-neural network method for c...   |  Next Patent:Chaotic signal producing method ->