An Enhanced EM algorithm using maximum entropy distribution as initial condition.
Guorong XuanYun Q. ShiPeiqi ChaiPatchara SutthiwanPublished in: ICPR (2012)
Keyphrases
- maximum entropy
- em algorithm
- expectation maximization
- maximum likelihood
- mixture model
- mixture distribution
- maximum entropy principle
- generative model
- probability density function
- parameter estimation
- maximum likelihood estimation
- gaussian mixture model
- likelihood function
- maximum likelihood estimates
- conditional random fields
- hyperparameters
- gaussian mixture
- incomplete data
- gaussian distribution
- log likelihood
- maximum a posteriori
- random variables
- hidden variables
- probability distribution
- density estimation
- bayesian framework
- unsupervised learning
- normal distribution
- training data
- exponential family
- expectation maximisation
- parameter learning
- text classification
- language model
- information extraction
- objective function