beta-Bernoulli process construction

E1031260

Bayesian nonparametric model feature allocation model latent feature model

The beta-Bernoulli process construction is a Bayesian nonparametric framework that generates sparse, infinite binary feature allocations by combining a beta process prior with Bernoulli-distributed feature indicators.

Jump to: Statements Referenced by

Statements (47)

Predicate	Object
instanceOf	Bayesian nonparametric model ⓘ feature allocation model ⓘ latent feature model ⓘ
appliesTo	feature allocation problems ⓘ latent factor analysis ⓘ matrix factorization with unknown number of features ⓘ multi-label modeling ⓘ nonparametric latent feature models for relational data ⓘ unsupervised learning ⓘ
assumes	exchangeability of objects ⓘ independent Bernoulli draws given beta process weights ⓘ
basedOn	completely random measures ⓘ
belongsTo	Bayesian nonparametrics NERFINISHED ⓘ probabilistic machine learning ⓘ
controlsFeatureReuseVia	concentration parameters of beta process ⓘ
controlsSparsityVia	mass parameter of beta process ⓘ
encourages	sparse feature allocations ⓘ
generalizes	finite latent feature models to infinite case ⓘ
generates	infinite binary feature allocations ⓘ
hasComponent	Bernoulli feature indicators per observation ⓘ beta process draw over feature weights ⓘ
hasGoal	flexible modeling of overlapping structure in data ⓘ
hasProperty	allows unbounded number of features ⓘ each observation has a sparse subset of features ⓘ features are shared across observations ⓘ
isAlternativeTo	Dirichlet process mixture models for clustering ⓘ
isDefinedOver	space of binary feature matrices ⓘ
isFormulatedIn	measure-theoretic probability framework ⓘ
isFrameworkFor	Bayesian nonparametric feature modeling ⓘ latent binary feature discovery ⓘ
isRelatedTo	Indian buffet process culinary metaphor ⓘ Indian buffet process stick-breaking construction NERFINISHED ⓘ
isUsedIn	Bayesian nonparametric clustering with overlapping clusters ⓘ Bayesian nonparametric factor models ⓘ latent feature topic models ⓘ nonparametric dictionary learning ⓘ nonparametric regression with latent features ⓘ
isUsedTo	infer number of latent features from data ⓘ
models	countably infinite set of features ⓘ
providesDeFinettiRepresentationFor	Indian buffet process NERFINISHED ⓘ
relatedTo	Indian buffet process NERFINISHED ⓘ
supportsInferenceVia	Gibbs sampling over feature indicators GENERATED ⓘ Markov chain Monte Carlo GENERATED ⓘ variational inference GENERATED ⓘ
usesLikelihood	Bernoulli distribution NERFINISHED ⓘ
usesPrior	beta process ⓘ
yields	binary feature matrix ⓘ

Referenced by (1)

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

Stick-breaking construction for the Indian buffet process → usesConcept → beta-Bernoulli process construction ⓘ