Show simple item record

Online Learning of Non-stationary Sequences

dc.date.accessioned2004-10-20T20:32:01Z
dc.date.accessioned2018-11-24T10:23:07Z
dc.date.available2004-10-20T20:32:01Z
dc.date.available2018-11-24T10:23:07Z
dc.date.issued2003-06-12en_US
dc.identifier.urihttp://hdl.handle.net/1721.1/7107
dc.identifier.urihttp://repository.aust.edu.ng/xmlui/handle/1721.1/7107
dc.description.abstractWe consider an online learning scenario in which the learner can make predictions on the basis of a fixed set of experts. The performance of each expert may change over time in a manner unknown to the learner. We formulate a class of universal learning algorithms for this problem by expressing them as simple Bayesian algorithms operating on models analogous to Hidden Markov Models (HMMs). We derive a new performance bound for such algorithms which is considerably simpler than existing bounds. The bound provides the basis for learning the rate at which the identity of the optimal expert switches over time. We find an analytic expression for the a priori resolution at which we need to learn the rate parameter. We extend our scalar switching-rate result to models of the switching-rate that are governed by a matrix of parameters, i.e. arbitrary homogeneous HMMs. We apply and examine our algorithm in the context of the problem of energy management in wireless networks. We analyze the new results in the framework of Information Theory.en_US
dc.format.extent48 p.en_US
dc.format.extent1815576 bytes
dc.format.extent911860 bytes
dc.language.isoen_US
dc.subjectAIen_US
dc.subjectonline learningen_US
dc.subjectrelative loss boundsen_US
dc.subjectswitching dynamicsen_US
dc.subjectwirelessen_US
dc.subject802.11en_US
dc.titleOnline Learning of Non-stationary Sequencesen_US


Files in this item

FilesSizeFormatView
AITR-2003-011.pdf911.8Kbapplication/pdfView/Open
AITR-2003-011.ps1.815Mbapplication/postscriptView/Open

This item appears in the following Collection(s)

Show simple item record