Pattern Recognition

Pattern recognition applications, feature extraction for patterns, data preprocessing, distance classifiers, decision functions, polynomial classifiers, clustering methods, self-organizing maps, Bayes classifiers, Maximum Likelihood methods, probabilistic inference, VC dimension, decision trees and random forests, perceptron, support vector machines.

At the end of the module students are able to apply different pattern recognition methods to a range of everyday and scientific problems. They are are able to analyse feature extraction and selection methods. They are able to analyse supervised und unsupervised classification methods, including the training of classifiers with machine learning techniques.

Basic linear algebra; for the exercises, rudimentary programming skills, ideally in Matlab; basic knowledge in statistics and signal representation.

Learning method:
In addition to the individual methods of the students, lecture contents are repeated and student understanding is facilitated by practical application in exercises. Exercise sheets are provided in advance of the respective tutorial session and should be solved a s (non-mandatory, ungraded) homework; this includes short programming tasks where Matlab templates are provided.

Teaching method:
During the lectures students are instructed in a teacher-centered style.
The exercises are held in a student-centered way.

The following kinds of media are used:
- Presentations with projected slides
- Lecture and tutorial notes
- Downloadable exercises with solutions

The following literature is recommended:
- R. O. Duda, P. E. Hart, D. G. Stork, Pattern Classification, 2.Auflage, John Wiley & Sons, 2001.
- C. Bishop, Pattern Recognition and Machine Learning, Springer, 2007

In a written exam (75 min) without aids students prove by answering short questions and by performing calculations that they are able to handle feature extraction methods, probabilistic inference, and machine learning techniques.