On free energy barriers in Gaussian priors and failure of MCMC for high-dimensional unimodal distributions.
Afonso S. BandeiraAntoine MaillardRichard NicklSven WangPublished in: CoRR (2022)
Keyphrases
- free energy
- posterior distribution
- high dimensional
- dirichlet process
- gaussian distribution
- probability distribution
- heavy tailed
- metropolis hastings algorithm
- markov chain monte carlo
- bayesian framework
- approximate inference
- expectation propagation
- variational bayes
- marginal distributions
- maximum a posteriori
- parameter estimation
- maximum likelihood
- bethe free energy
- latent variables
- belief propagation
- hyperparameters
- random variables
- gaussian process
- low dimensional
- graphical models
- exponential family
- posterior probability
- feature space
- expectation maximization
- fixed point
- gaussian mixture model
- competitive learning
- dimensionality reduction
- data points
- probability density function
- parameter space
- joint distribution
- model selection
- markov chain
- markov random field
- upper bound
- generative model
- neural network