A low-cost variational-Bayes technique for merging mixtures of probabilistic principal component analyzers.
Pierrick BruneauMarc GelgonFabien PicarougnePublished in: Inf. Fusion (2013)
Keyphrases
- principal components
- variational bayes
- latent variables
- probabilistic model
- bayesian inference
- principal component analysis
- hyperparameters
- expectation maximization
- mixture model
- gaussian mixture model
- bayesian learning
- latent dirichlet allocation
- free energy
- dimensionality reduction
- generative model
- bayesian networks
- exponential family
- log likelihood
- model selection
- em algorithm
- posterior distribution
- data sets
- information theoretic
- maximum likelihood
- feature set
- regression model
- upper bound
- incremental learning
- feature space
- closed form
- covariance matrix
- support vector
- topic models