Pattern Recognition - Teoria (2008/2009)

Learning outcomes

The Pattern Recognition course aims at describing the theoretical foundations and the main methods related to the analysis and automatic recognition of data or "pattern". This discipline is at the basis and complete many other, more diffuse, disciplines like image processing, computer vision, artificial intelligence, data mining, databases, and others.

During the course, enphasis will be posed on probabilitsic techniques with particular reference to the learning of systems devoted to recognition (of, but not exclusively, images) and neural netowrks.

The involved applications are many and diverse. Among the others already quoted we can cite bioinformatics, analysis and interpretation of biomedical and biological data (e.g., genomics, proteomics, etc.), biometry, videosurveillance, robotics, speech recognition.

* Introduction: what, what is useful for, systems, applications.
* Recognition and classification.
* Feature extraction and representation.
* Bayes theory.
* Parameter estimation: parametric and non parametric methods.
* Linear classifier, nonlinear classifier and discriminant functions.
* Structural pattern recognition (just a sketch).
* Feature selection.
* Neural networks.
* Clustering.
* Advanced methoids: Hidden Markov Models.

The course is structured in 32 hours of theory and 12 hours of laboratory activities (5 credits). Lab activities consist of practice and problem solving using MATLAB.

The exam consists of:
- a project,
- short oral interview.
The project will involve one or more topics discussed in the lectures, with reference to image/video analysis and bioinformatics problems.
The oral interview will check the knowledge of the lectures' topics and can be substituted by a written examination with a few short questions likewise the oral one.