Salvatore Ingrassia University of Catania (Italy)
Date:
Event location: Aula IV, Department of Statistical Sciences, Via Belle Arti 41 (STAT PhD Classes Virtual Room on need)
Type: Cycle 41 - Elective Courses
Short course on Learning from Data via Mixture Models
Prof. Salvatore Ingrassia, University of Catania (Italy), salvatore.ingrassia@unict.it
The teaching program concerns Mixture Models and their applications. In detail, the following main topics will be presented:
1. Mixtures of distributions.
a. Gaussian mixtures, Model-Based Clustering, Maximum Likelihood approach to parameter estimation and numerical issues, EM algorithm.
b. Measures of Class Agreement: Confusion matrix, Adjusted Rand Index(ARI).
c. Parsimonious model-based clustering.
2. Mixtures of regressions.
a. Mixtures of regression with fixed covariates, mixtures of regression with concomitant variables.
b. Cluster-weighted models, mixtures of generalized linear models.
c. Local and overall R-squared measures for mixtures of regression models.
d. Conditional mixture modeling and model-based clustering.
e. Mixtures of experts.
3. Skew Distributions and Related Mixture Models
a. Skew Distributions.
b. Skew-normal and skew-t mixtures.
4. Transformation Mixture Models
a. Manly transformation for normality.
b. Bickel-Doksum transformation for normality.
Teaching schedule: 10 hrs.