Training RBMs based on the signs of the CD approximation of the log-likelihood derivatives.
Asja FischerChristian IgelPublished in: ESANN (2011)
Keyphrases
- log likelihood
- restricted boltzmann machine
- feed forward neural networks
- closed form
- information theoretic
- scoring function
- density estimation
- em algorithm
- deep learning
- maximum likelihood
- higher order
- back propagation
- artificial neural networks
- training set
- learning algorithm
- neural network
- conditional random fields
- probabilistic graphical models
- computer vision
- generative model
- supervised learning
- semi supervised
- hidden layer
- probabilistic model