Expected likelihood approach for low sample support covariance matrix estimation in angular central Gaussian distributions.
Olivier BessonYuri I. AbramovichPublished in: ACSSC (2013)
Keyphrases
- covariance matrix
- covariance matrices
- multivariate gaussian
- gaussian distribution
- maximum likelihood
- maximum likelihood estimation
- sample size
- estimation error
- normal distribution
- mixture of gaussians
- gaussian mixture
- principal component analysis
- expectation maximization
- pseudo inverse
- model selection
- gaussian mixture model
- mahalanobis distance
- class conditional densities
- eigenvalues and eigenvectors
- computer vision
- parameter estimation
- mixture model
- em algorithm
- probabilistic model
- optimal solution
- density estimation
- similarity search
- noise model
- log likelihood
- kl divergence
- correlation matrix
- geometrical interpretation
- feature extraction