An HDP-HMM for Systems With State Persistence

Simple item page
dc.contributor.authorFox, Emily B
dc.contributor.authorSudderth, Erik B
dc.contributor.authorJordan, Michael I
dc.contributor.authorWillsky, Alan S
dc.date2023-05-17T15:27:31.000
dc.date.accessioned2023-05-23T03:33:47Z
dc.date.available2023-05-23T03:33:47Z
dc.date.issued2008-01-01
dc.date.submitted2016-08-19T13:04:17-07:00
dc.description.abstractThe hierarchical Dirichlet process hidden Markov model (HDP-HMM) is a flexible, nonparametric model which allows state spaces of unknown size to be learned from data. We demonstrate some limitations of the original HDP-HMM formulation (Teh et al., 2006), and propose a sticky extension which allows more robust learning of smoothly varying dynamics. Using DP mixtures, this formulation also allows learning of more complex, multimodal emission distributions. We further develop a sampling algorithm that employs a truncated approximation of the DP to jointly resample the full state sequence, greatly improving mixing rates. Via extensive experiments with synthetic data and the NIST speaker diarization database, we demonstrate the advantages of our sticky extension, and the utility of the HDP-HMM in real-world applications.
dc.identifier.urihttps://repository.upenn.edu/handle/20.500.14332/47467
dc.legacy.articleid1486
dc.legacy.fulltexturlhttps://repository.upenn.edu/cgi/viewcontent.cgi?article=1486&context=statistics_papers&unstamped=1
dc.source.beginpage312
dc.source.endpage319
dc.source.issue116
dc.source.journalStatistics Papers
dc.source.journaltitleProceedings of the 25th International Conference on Machine Learning
dc.source.statuspublished
dc.subject.otherStatistics and Probability
dc.titleAn HDP-HMM for Systems With State Persistence
dc.typePresentation
digcom.contributor.authorFox, Emily B
digcom.contributor.authorSudderth, Erik B
digcom.contributor.authorJordan, Michael I
digcom.contributor.authorWillsky, Alan S
digcom.identifierstatistics_papers/116
digcom.identifier.contextkey9005594
digcom.identifier.submissionpathstatistics_papers/116
digcom.typeconference
dspace.entity.typePublication
upenn.schoolDepartmentCenterStatistics Papers
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
HDP_HMM_systems.pdf
Size:
494.63 KB
Format:
Adobe Portable Document Format
Download
Collection
Presentations