Pitman–Yor process models

E1031256

Bayesian nonparametric model probabilistic model

Pitman–Yor process models are Bayesian nonparametric models that generalize Dirichlet process models by allowing power-law behavior and heavier-tailed distributions over clusters.

Try in SPARQL Jump to: Statements Referenced by

Statements (47)

Predicate	Object
instanceOf	Bayesian nonparametric model ⓘ probabilistic model ⓘ
allow	heavier-tailed distributions over clusters ⓘ power-law behavior in cluster sizes ⓘ
appliedTo	power-law phenomena in data ⓘ vocabulary growth modeling ⓘ word frequency modeling ⓘ
assume	exchangeability of observations ⓘ
basedOn	Pitman–Yor process NERFINISHED ⓘ
canBeExtendedTo	hierarchical Pitman–Yor process models ⓘ
characterizedBy	exchangeable partition probability function ⓘ power-law distribution over number of clusters ⓘ rich-get-richer clustering behavior ⓘ
comparedTo	Dirichlet process models NERFINISHED ⓘ
enable	flexible prior over partitions ⓘ modeling of heavy-tailed cluster size distributions ⓘ
extend	Dirichlet process mixture models NERFINISHED ⓘ
generalize	Chinese restaurant process models ⓘ Dirichlet process models NERFINISHED ⓘ
hasHyperparameter	base distribution ⓘ concentration parameter ⓘ discount parameter ⓘ
implementedIn	Markov chain Monte Carlo inference methods NERFINISHED ⓘ sequential Monte Carlo methods ⓘ variational inference methods ⓘ
mathematicallyBasedOn	random partitions ⓘ stable subordinators ⓘ
originatedFrom	work of Jim Pitman ⓘ work of Marc Yor ⓘ
provide	more flexible prior over cluster sizes than Dirichlet process models ⓘ
relatedTo	Chinese restaurant process NERFINISHED ⓘ Dirichlet process NERFINISHED ⓘ Poisson–Dirichlet distribution NERFINISHED ⓘ
specialCaseOf	two-parameter Poisson–Dirichlet process models ⓘ
support	countably infinite number of clusters ⓘ power-law tails in predictive distributions ⓘ
usedFor	clustering ⓘ density estimation ⓘ hierarchical Bayesian modeling ⓘ language modeling ⓘ mixture modeling ⓘ nonparametric prior over discrete distributions ⓘ topic modeling ⓘ
usedIn	Bayesian statistics NERFINISHED ⓘ machine learning ⓘ natural language processing ⓘ nonparametric Bayesian inference ⓘ

Referenced by (1)

Full triples — surface form annotated when it differs from this entity's canonical label.

Dirichlet process models → relatedTo → Pitman–Yor process models ⓘ