Coursebooks 2017-2018

PDF
 

Statistical Sequence Processing

EE-605

Lecturer(s) :

Bourlard Hervé

Language:

English

Frequency

Every 2 years

Remarque

Every 2 years. Next time: Spring 2019

Summary

This course discusses advanced methods extensively used for the processing, prediction, and classification of temporal (multi-dimensional and multi-channel) sequences. In this context, it also describes key links between signal processing, linear algebra, statistics and artificial neural networks.

Content

1. Introduction: Statistical (static and dynamic) pattern recognition, temporal pattern recognition problems, etc
2. Basic tools in temporal pattern modeling: Correlation, autocorrelation, linear/nonlinear AR, ARMA and ARCH modeling
3. Statistical pattern recognition: Bayes classifiers, artificial neural networks (ANNs), discriminant functions, Expectation-Maximization algorithm, dynamic programming
4. Sequence processing: discrete Markov models, hidden Markov models (HMM), autoregressive (AR)-HMM, hybrid HMM/ANN systems, parameter estimation (EM and forward-backward algorithms applied to these models)
5. Laboratory exercises: in statistical pattern recognition, autoregressive modeling, Markov models and hidden Markov models

Note

Course notes (and relevant book chapters) available.

Keywords

Statistical modeling, Markov models, hidden Markov models, artificial neural networks for sequence processing.

Learning Prerequisites

Recommended courses

Undergraduate level statistics, linear algebra (matric computations, up to PCA) and minimum knowledge/interest in signal processing and machine learning. Programming in Matlab or similar.

Assessment methods

Multiple.

Resources

Websites

In the programs

Reference week

 
      Lecture
      Exercise, TP
      Project, other

legend

  • Autumn semester
  • Winter sessions
  • Spring semester
  • Summer sessions
  • Lecture in French
  • Lecture in English
  • Lecture in German