Stable Training of Probabilistic Models Using the Leave-One-Out Maximum Log-Likelihood Objective.
Kutay BölatSimon H. TindemansPeter PalenskyPublished in: CoRR (2023)
Keyphrases
- log likelihood
- probabilistic model
- information theoretic
- density estimation
- maximum likelihood
- scoring function
- graphical models
- training set
- mixture model
- exponential family
- expectation maximization
- em algorithm
- penalized likelihood
- latent variables
- generative model
- bayesian networks
- bayesian inference
- conditional random fields
- cross validation
- mutual information
- hidden variables
- supervised learning
- feature vectors
- similarity measure