{"title":"On asymptotically optimal sequential lossless coding for memoryless switching sources","authors":"G. Shamir","doi":"10.1109/ISIT.2002.1023396","DOIUrl":null,"url":null,"abstract":"Two strategies are used to exploit a set D/sub n/ of dormant nodes. The first creates /spl sigma/ nodes each time a new node should be created. One node with reset occurrence counts, while the others assume counts of all dormant nodes. When sufficient data is available, all new nodes but the one that is likely to represent a true state are eliminated. The second strategy creates only one instance, but when sufficient data is available, the occurrence counts of the new node are changed to those of the most likely dormant state, if such exists. If some mild regularity conditions hold, and only the most recent /spl sigma/ states may repeat, both strategies are asymptotically optimal. In practice, the strategies vary in complexity and performance.","PeriodicalId":92129,"journal":{"name":"Proceedings. IEEE International Symposium on Information Theory","volume":"1 1","pages":"124-"},"PeriodicalIF":0.0000,"publicationDate":"2002-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings. IEEE International Symposium on Information Theory","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ISIT.2002.1023396","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 2
Abstract
Two strategies are used to exploit a set D/sub n/ of dormant nodes. The first creates /spl sigma/ nodes each time a new node should be created. One node with reset occurrence counts, while the others assume counts of all dormant nodes. When sufficient data is available, all new nodes but the one that is likely to represent a true state are eliminated. The second strategy creates only one instance, but when sufficient data is available, the occurrence counts of the new node are changed to those of the most likely dormant state, if such exists. If some mild regularity conditions hold, and only the most recent /spl sigma/ states may repeat, both strategies are asymptotically optimal. In practice, the strategies vary in complexity and performance.