Training restricted Boltzmann machines using approximations to the likelihood gradient.
Tijmen TielemanPublished in: ICML (2008)
Keyphrases
- restricted boltzmann machine
- contrastive divergence
- deep learning
- conditional random fields
- deep belief networks
- learning algorithm
- probabilistic graphical models
- maximum likelihood
- optical flow
- training set
- machine learning
- hidden markov models
- collaborative filtering
- unsupervised learning
- hidden layer
- free energy