LEARNING OUTCOMES

Analysis and visualization of data. Static graphical models (Bayesian networks, regression, factor analysis, decision theory and Gaussian mixtures). Dynamic graphic models (hidden markov models, linear dynamic models). Neural networks. Vector quantization and vector support machines. Bayesian approach to model comparison and hypothesis testing.

AIMS AND LEARNING OUTCOMES

The aim of this course is to provide students with the essential tools for quantitative analysis of data and signals of interest for medicine and biology.

By the end of the course the students will be able to:

1. Design and apply methods of analysis and modelling of data - including temporal data (signals) - of interest for medicine and biology

2. Identify the correct approach (model selection, model identification, data visualization) for a specific data analysis problem

3. Use MATLAB to correctly and effectively visualizing biomedical data and signals and to implement models of them

PREREQUISITES

There are no formal prerequisites, but  the course requires solid foundations in mathematical analysis and linear algebra.

TEACHING METHODS

The course is organised as a combination of lectures and lab activities. 

Lectures will focus on theory and methods for data analysis. 

Lab activities will focus on application to real data analysis problems in the context of bioengineering.

In AY 2020-2021 the course will be taught in presence but all lectures and lab activities will be also offered at a distance, both online and offline.

SYLLABUS/CONTENT

A. Data Analysis and Data visualization: Types of data. Analysis as modeling. Statistical data analysis. The do’s and don’t’s of data visualization

B. Probability density estimates: Unsupervised learning. Gaussian model. Principal Component Analysis, Factor Analysis, Independent Component Analysis, Cluster analysis and the EM algorithm.

C. Pattern analysis and decision theory: Bayesian decision theory. Bayes classifiers. Logistic classifiers. Generalized linear models. Performance of a classifier: ROC curve. 

D. Multilayer neural networks: Perceptrons. Multilayer neural networks and the backpropagation algorithm. Design of a neural network model

E. Graphical Models: Static data. A general framework for data modeling. The EM algorithm. Regression, decision theory, factor analysis as graphical models. Dynamic graphical models (overview): Temporal data (signals). Discrete vs continuous signals. Hidden Markov Models, Linear dynamical systems as dynamic graphical models

F. Generalization: Regularization theory, Vector Quantization. Support Vector Machines.

RECOMMENDED READING/BIBLIOGRAPHY

Murphy, KP. Machine Learning: A Probabilistic Perspective.  MIT Press, 2012.