{"title":"Construction of multi-layer feedforward binary neural network by a genetic algorithm","authors":"C. Chow, Tong Lee","doi":"10.1109/IJCNN.2002.1007547","DOIUrl":null,"url":null,"abstract":"An approach is introduced to determine the topology of a feedforward binary neural network automatically. The approach is based on a construction algorithm that constructs one layer of hidden nodes at a time until the problem is solved. In each layer, the algorithm determines the necessary number of nodes through a growth process by finding the best hidden node that would help to partition the input training data set. This is done using a genetic algorithm. The proposed algorithm can determine the necessary number of hidden layers and number of hidden nodes at each layer automatically. Tests on a number of benchmark problems illustrated the effectiveness of the proposed technique, both in terms of network complexity and recognition accuracy, compared with a geometrical learning approach.","PeriodicalId":382771,"journal":{"name":"Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290)","volume":"83 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2002-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"3","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/IJCNN.2002.1007547","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 3
Abstract
An approach is introduced to determine the topology of a feedforward binary neural network automatically. The approach is based on a construction algorithm that constructs one layer of hidden nodes at a time until the problem is solved. In each layer, the algorithm determines the necessary number of nodes through a growth process by finding the best hidden node that would help to partition the input training data set. This is done using a genetic algorithm. The proposed algorithm can determine the necessary number of hidden layers and number of hidden nodes at each layer automatically. Tests on a number of benchmark problems illustrated the effectiveness of the proposed technique, both in terms of network complexity and recognition accuracy, compared with a geometrical learning approach.