Wide Bayesian neural networks have a simple weight posterior: theory and accelerated sampling.
Jiri HronRoman NovakJeffrey PenningtonJascha Sohl-DicksteinPublished in: CoRR (2022)
Keyphrases
- neural network
- posterior distribution
- markov chain monte carlo
- posterior probability
- pattern recognition
- genetic algorithm
- artificial neural networks
- theoretical framework
- back propagation
- multi layer
- probability distribution
- wide range
- bayesian networks
- data sets
- random sampling
- bayesian framework
- sampling strategy
- monte carlo
- computational model
- parameter estimation
- em algorithm
- maximum likelihood
- graphical models
- fuzzy logic
- probabilistic model