This course will provide an introduction to modern techniques for Bayesian pattern recognition and machine learning. We will begin with a review of linear regression, generative and discriminative methods for linear classification and the use of Kernel Methods and Support Vector Machines. We will then examine Bayesian Networks, Gaussian Mixture Models and principal components analysis.

The course will provide basic competence in machine learning methods that can be used to construct systems for data mining, communications, signal analysis, computer vision, speech recognition, man-machine interaction, and intelligent systems.

This course will be taught in French using an english language text book.

The course will closely follow the development of the text book by Chris Bishops text book on Pattern Recognition and Machine Learning.

S1: Introduction, probability distributions, likelihood.
S2: Bayesian Probability Theory
S3: Gaussian Probability Density Functions
S4: Multivariate Gaussian Probability Density Functions
S5: Principal Component Analysis
S6: Mixture Models and Clustering
S7: Generative Methods for Classification
S8: Fisher Linear Discriminant and Perceptrons
S9: Perceptron and Kernel Methods
S10: Neural Networks
S11: Support Vector Machines
S12: Combining Models, Boosting.

Linear Algebra,
Multivariate Calculus
Optimization
Probability and Statistics.

Purchase, or access to the following textbook is STRONGLY recommended:
Christopher M. Bishop, Pattern Recognition and Machine Learning, Springer Verlag, 2006.

Both Session 1 Exam and the make-up exam will be 3 hour written exams with documents authorised.

N1=E1
N2=E2