Explaining and Generalizing Skip-Gram through Exponential Family Principal Component Analysis.
Ryan CotterellAdam PoliakBenjamin Van DurmeJason EisnerPublished in: EACL (2) (2017)
Keyphrases
- exponential family
- principal component analysis
- maximum likelihood
- log likelihood
- graphical models
- closed form
- statistical models
- density estimation
- mixture model
- missing values
- dimensionality reduction
- low dimensional
- face recognition
- order statistics
- feature extraction
- hidden variables
- probability density function
- feature space
- variational methods
- probabilistic model
- high dimensional data
- markov chain monte carlo
- information theoretic
- statistical model
- belief propagation
- parameter estimation
- probabilistic inference
- markov random field
- mutual information