Learning outcomes

Pattern Recognition (PR) focuses on creating classifiers, that is, algorithms that can learn aspects of the reality and make appropriate decisions once in the presence of new stimuli. Examples of classifiers are: speech recognition, automotive applications, surveillance systems, quality control systems, recommendation systems. The PR course intends to provide the methodological principles at the basis of the classification, together with the most modern techniques that can solve problems which were unmanageable until a few years ago. At the end of the course, the student will have to demonstrate how to solve a classification problem by applying the most suitable instrument to the case, justifying the theoretical choices.

Program

The course program can be divided into two parts, the methodological and the applicative one, which will go hand in hand with the lessons.

Methodologies
- Introduction: what is the classification, classification systems, and classification applications
- Statistical classification
- Bayesian decision theory
- Linear, non-linear and discriminative machines
--Selection and extraction of features, PCA and Fisher transform
--Parametric classifiers and how to train them
-- Expectation-Maximization Algorithm and Gaussian Mixtures
-- Non parametric classifiers and how to train them
--Hidden Markov Models
--Unsupervised classification methods (clustering)


Applications
-- Binary and multiclass classification on real benchmarks
-- Faces recognition
--Tracking


Textbooks:
- Richard O. Duda, Peter E. Hart, and David G. Stork. 2000. Pattern Classification (2nd Edition). Wiley-Interscience.
- Christopher M. Bishop. 2006. Pattern Recognition and Machine Learning (Information Science and Statistics). Springer-Verlag New York, Inc., Secaucus, NJ, USA.

Examination Methods

Examination method is oral; the required content will be those seen during the lessons, as indicated by the course program. In particular, when necessary, a formal demonstration of a procedure will be requested. In all cases, the questions will address a classification problem where the student will have to suggest the most suitable technique for the case, formally demonstrating the choice. The final vote will be built depending on the student's proposed solution to the question (20 points total), and the formal accuracy with which the solution is presented (10 points).

Teaching materials e documents

•   Lab.1 Bayes Classifiers   (zip, it, 6506 KB, 3/22/17)
•   Lab.2 PCA   (zip, it, 2 KB, 4/3/17)
•   Lab.3 Eigenfaces   (zip, it, 5046 KB, 4/5/17)
•   Lab.4 Fisher Discriminant Analysis   (zip, it, 2 KB, 4/19/17)
•   Lab.5 Fisherfaces   (zip, it, 5045 KB, 4/19/17)
•   Lab.6 EM   (zip, it, 16056 KB, 5/8/17)
•   Lab.7 Stima non parametrica di densità - tracking   (zip, it, 5 KB, 5/24/17)
•   Lez.0 Introduzione al corso   (zip, it, 514 KB, 3/6/17)
•   Lez.1 Paradigma di Bayes   (zip, it, 14941 KB, 3/6/17)
•   Lez.2 Feature Extraction - PCA e FLDA   (zip, it, 1698 KB, 4/26/17)
•   Lez.3 ML estimation and EM algorithm   (zip, it, 13228 KB, 5/2/17)
•   Lez.4 Stima non parametrica e tracking   (zip, it, 5773 KB, 5/10/17)
•   Lez.5 Classificazione non supervisionata   (zip, it, 2668 KB, 5/24/17)