Morfessor EM+Prune: Improved Subword Segmentation with Expectation Maximization and Pruning.
Stig-Arne GrönroosSami VirpiojaMikko KurimoPublished in: CoRR (2020)
Keyphrases
- expectation maximization
- em algorithm
- image segmentation
- search space
- expectation maximisation
- mixture model
- probabilistic model
- segmentation algorithm
- maximum likelihood
- segmentation method
- unsupervised learning
- level set
- maximum likelihood estimation
- maximum a posteriori
- generative model
- gaussian mixture model
- bayesian framework
- shape prior
- gaussian mixture
- pruning strategy
- segmentation accuracy
- parameter estimation
- n gram
- gradient ascent
- medical images
- gaussian distribution
- density estimation
- probability density function
- computer vision
- image analysis
- energy function
- pruning algorithm
- broadcast news
- log likelihood function
- region growing
- fully automatic
- image regions
- test collection
- speech recognition
- graph cuts
- multiscale