Michele Casoni(University of Siena)

When: Apr 13th, 2022 – 11:00 – 11:45 AM

Where: Google meet link

Description

Mixture models

A well known problem in the unsupervised learning framework is how to approximate an empirical distribution by a probability density function.
A way for approaching it is provided by Mixture Models.
In these probabilistic models, the empirical distribution is approximated by a linear combination of probability density functions, which is called a Mixture. The parameters of a Mixture can be found applying the Expectation – Maximization (EM) algorithm.

In this seminar, I will describe my personal implementation of the EM algorithm, which generalizes the usual Matlab libraries making use of non-Gaussian components for a Mixture.