BAYESIAN NONPARAMETRICS BILBLIOGRAPHY (Focus: Hierarchical and sequential Modelling, Supervision and applications to Videos)


Hierarchical Dirichlet Processes Teh, Jordan, Beal and Blei, Journal of American Statistical Association (JASA), 2006

Hierarchical Bayesian model with infinitely many measures shared across different pre-defined groups of data, no sequential ordering considered (Group Exchangeable)

An HDP-HMM for Systems with State Persistence Fox, Sudderth, Jordan and Willsky, ICML, 2008

A mixture model with infinitely many mixture components, with transition probabilities from one component to another, self-transitions favoured, considers sequential ordering of data (MArkov Exchangeability)

Infinite Latent Feature Models and the Indian Buffet Process Griffiths and Grahramani, NIPS 2005

Feature-sharing by datapoints, no sequential ordering

Hierarchical Beta Processes and the Indian Buffet Process Thibaux and Jordan, AISTATS 2007

Theoretical treatment of above

THE NESTED DIRICHLET PROCESS Rodriguez, Dunson and Gelfand, Journal of American Statistical Association (JASA), 2008

2-level hierarchy of measures, pre-defined groups of data share measures over measures, no sequential ordering of data considered

A Hierarchical Dirichlet Process Model with Multiple Levels of Clustering for Human EEG Seizure Modeling Wulsin, Jensen and Litt, ICML 2012

HDP extended to 3-levels, implies NDP-like hierarchy of measures, no sequential ordering of data

The Hierarchical Dirichlet Process Hidden Semi-Markov Model Johnson and Willsky, UAI 2011

Similar to HDP-HMM but state change probability depends on the duration of current state

Distance Dependent Chinese Restaurant Processes Blei and Frezier, ICML 2011

Considers "distances" between different datapoints in assigning them to mixture components

Dirichlet Process with Mixed Random Measures: A Nonparametric Topic Model for Labeled Data Kim, Kim and Oh, ICML 2012

Similar to HDP but allows biasing of the predefined groups towards certain mixture components

Effective Split-Merge Monte Carlo Methods for Nonparametric Models of Sequential Data Hughes, Fox and Sudderth, NIPS 2012
Split-merge inference for BP-HMM, with sharing of features among multiple sequences of data

Bayesian Nonparametrics for Videos

Multi-Class Video Co-Segmentation with a Generative Multi-Video Model Chiu and Fritz, CVPR 2013
Makes use of DDCRP.

Nonparametric discovery of activity patterns from video collections Hughes and Sudderth, CVPRW 2012
Makes use of Beta Process HMMs

Extracting and Locating Temporal Motifs in Video Scenes Using a Hierarchical Non Parametric Bayesian Model Emonet, Varadarajan and Odobez, CVPR 2011
Makes use of a HDP-like model with timestamps