Sparse adaptive dirichlet-multinomial-like processes

dc.contributor.authorHutter, Marcusen
dc.date.accessioned2025-12-31T18:41:37Z
dc.date.available2025-12-31T18:41:37Z
dc.date.issued2013en
dc.description.abstractOnline estimation and modelling of i.i.d. data for short sequences over large or complex "alphabets" is a ubiquitous (sub)problem in machine learning, information theory, data compression, statistical language processing, and document analysis. The Dirichlet-Multinomial distribution (also called Polya urn scheme) and extensions there of are widely applied for online i.i.d. estimation. Good a-priori choices for the parameters in this regime are difficult to obtain though. I derive an optimal adaptive choice for the main parameter via tight, data-dependent redundancy bounds for a related model. The 1-line recommendation is to set the 'total mass' = 'precision' = 'concentration' parameter to m/[2 ln n+1/m], where n is the (past) sample size and m the number of different symbols observed (so far). The resulting estimator is simple, online, fast, and experimental performance is superb.en
dc.description.statusPeer-revieweden
dc.format.extent28en
dc.identifier.issn1532-4435en
dc.identifier.scopus84898021490en
dc.identifier.urihttps://hdl.handle.net/1885/733797761
dc.language.isoenen
dc.relation.ispartofseries26th Conference on Learning Theory, COLT 2013en
dc.sourceJournal of Machine Learning Researchen
dc.subjectAdaptive parametersen
dc.subjectData compressionen
dc.subjectData-dependent redundancy bounden
dc.subjectDirichlet-Multinomialen
dc.subjectPolya urnen
dc.subjectSmall/large alphabeten
dc.subjectSparse codingen
dc.titleSparse adaptive dirichlet-multinomial-like processesen
dc.typeConference paperen
dspace.entity.typePublicationen
local.bibliographicCitation.lastpage459en
local.bibliographicCitation.startpage432en
local.contributor.affiliationHutter, Marcus; School of Computing, ANU College of Systems and Society, The Australian National Universityen
local.identifier.ariespublicationu4334215xPUB1166en
local.identifier.citationvolume30en
local.identifier.pure9c45ccef-8abd-4c99-9f9d-09e9fc8d0549en
local.identifier.urlhttps://www.scopus.com/pages/publications/84898021490en
local.type.statusPublisheden

Downloads