Reverse-time diffusion processes in discrete time

dc.contributor.authorDasgupta, Souraen
dc.contributor.authorAnderson, Brian D. O.en
dc.date.accessioned2026-02-04T12:41:10Z
dc.date.available2026-02-04T12:41:10Z
dc.date.issued2025en
dc.description.abstractGenerative AI relies on finding reverse models for linear discrete time forward diffusions with non-Gaussian initial states, but uses indirect methods for reversal as there is no theory for direct reversal in discrete time. We give sufficient conditions that guarantee the existence of a reverse time diffusion and give a method of meeting them. We also give a necessary and sufficient condition for the reverse model to be input-affine.en
dc.description.statusPeer-revieweden
dc.format.extent6en
dc.identifier.isbn979-8-3315-2627-6en
dc.identifier.otherdblp:conf/cdc/DasguptaA25en
dc.identifier.otherORCID:/0000-0002-1493-4774/work/204381149en
dc.identifier.urihttps://hdl.handle.net/1885/733805261
dc.language.isoenen
dc.publisherIEEEen
dc.relation.ispartofProceedings of 64th IEEE Conference on Decision and Controlen
dc.rights © 2025 IEEEen
dc.titleReverse-time diffusion processes in discrete timeen
dc.typeConference paperen
dspace.entity.typePublicationen
local.bibliographicCitation.lastpage5492en
local.bibliographicCitation.startpage5487en
local.contributor.affiliationDasgupta, Soura; University of Iowaen
local.contributor.affiliationAnderson, Brian D. O.; School of Engineering, ANU College of Systems and Society, The Australian National Universityen
local.identifier.doi10.1109/CDC57313.2025.11312217en
local.identifier.pure9fd3c61b-dcfd-4ff2-96db-80a7a276e89een
local.type.statusPublisheden

Downloads