Training Restricted Boltzmann Machine via the Thouless-Anderson-Palmer free energy.
Marylou GabriéEric W. TramelFlorent KrzakalaPublished in: NIPS (2015)
Keyphrases
- restricted boltzmann machine
- free energy
- loopy belief propagation
- belief propagation
- deep learning
- probabilistic graphical models
- approximate inference
- fixed point
- conditional random fields
- graphical models
- competitive learning
- upper bound
- learning algorithm
- graph cuts
- probabilistic model
- neural network
- markov networks
- energy minimization
- probabilistic inference
- markov random field
- training set