{"title":"具有类所属粒化的颗粒神经网络模型","authors":"D. A. Kumar, S. Meher","doi":"10.1109/IC3I.2014.7019743","DOIUrl":null,"url":null,"abstract":"Granular neural networks (GNNs) take the fuzzy granulated input and process them through neural networks (NNs). As a result, performance of GNNs depends highly on the granulation process and initial weights of NNs. The initial weights between nodes of GNNs provide the starting point in the searching of the lowest cost function value. The present article proposes GNN model that use class-belonging (CB) fuzzy granulation of input information and rough set-theoretic weight initialization of NNs. The model thus avoids the random initialization of weights and provides improved decisions at the output with CB granulation. Classification performance of the proposed GNN model has been assessed using various measurement indexes and its superiority over similar other methods is justified. Conventional back propagation algorithm is used to train the proposed model of GNN.","PeriodicalId":430848,"journal":{"name":"2014 International Conference on Contemporary Computing and Informatics (IC3I)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2014-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"5","resultStr":"{\"title\":\"Granular neural networks models with class-belonging granulation\",\"authors\":\"D. A. Kumar, S. Meher\",\"doi\":\"10.1109/IC3I.2014.7019743\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Granular neural networks (GNNs) take the fuzzy granulated input and process them through neural networks (NNs). As a result, performance of GNNs depends highly on the granulation process and initial weights of NNs. The initial weights between nodes of GNNs provide the starting point in the searching of the lowest cost function value. The present article proposes GNN model that use class-belonging (CB) fuzzy granulation of input information and rough set-theoretic weight initialization of NNs. The model thus avoids the random initialization of weights and provides improved decisions at the output with CB granulation. Classification performance of the proposed GNN model has been assessed using various measurement indexes and its superiority over similar other methods is justified. Conventional back propagation algorithm is used to train the proposed model of GNN.\",\"PeriodicalId\":430848,\"journal\":{\"name\":\"2014 International Conference on Contemporary Computing and Informatics (IC3I)\",\"volume\":\"7 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2014-11-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"5\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2014 International Conference on Contemporary Computing and Informatics (IC3I)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/IC3I.2014.7019743\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2014 International Conference on Contemporary Computing and Informatics (IC3I)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/IC3I.2014.7019743","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Granular neural networks models with class-belonging granulation
Granular neural networks (GNNs) take the fuzzy granulated input and process them through neural networks (NNs). As a result, performance of GNNs depends highly on the granulation process and initial weights of NNs. The initial weights between nodes of GNNs provide the starting point in the searching of the lowest cost function value. The present article proposes GNN model that use class-belonging (CB) fuzzy granulation of input information and rough set-theoretic weight initialization of NNs. The model thus avoids the random initialization of weights and provides improved decisions at the output with CB granulation. Classification performance of the proposed GNN model has been assessed using various measurement indexes and its superiority over similar other methods is justified. Conventional back propagation algorithm is used to train the proposed model of GNN.