|Website:||website containing additional information|
|Credits:||7.5 ECTS (=5.25 old credit points)|
|Period:||periode 4 (week 17 t/m 27, dwz 25-4-2005 t/m 8-7-2005; herkansing week 35)
|Participants:||up till now 19 subscriptions|
|Schedule:||Dit is een oud rooster!
|Contents:||The course Pattern Recognition is about the classification and analysis
As an important example we will look at patterns in images and music notation, but the methods are generally applicable.
There are numerous applications of pattern recognition techniques, such as
industrial inspection (e.g. quality control of materials), biomedical
inspection (e.g. chromosome analysis), remote sensing (earth observation),
astronomy (galaxy research), and security (e.g. fingerprint and handwriting
Because more and more measured data is generated, the need for automatic
analysis also increases.
We will look at patterns in two ways: as a collection of features (such as
color and image gradient direction) that occur with a certain probability,
and as a configuration of geometric primitives (such as points, lines,
The two corresponding ways of pattern recognition are statistical and
geometrical pattern recognition.
|Literature:||For statistical pattern recognition:
Duda, Hart, Stork: Pattern Classification (2nd ed), John Wiley 2001.
For geometrical pattern recognition:
articles will be made available.
After a few introductory lectures by the lecturer, about statistical pattern recognition, and geometrical shape recognition, students present specific articles
and book chapters.
|Exam form:||The grade depends on the given presentations (40%) summaries (40%) and a small exam (20%).
|Minimum effort to qualify for 2nd chance exam:||Om aan de aanvullende toets te mogen meedoen is ontbreken van ten hoogte 1 toetsactiviteit toegestaan.|
|Description:||Topics that are treated are: probability density functions, Bayesian
decision theory, feature space, supervised and unsupervised classification,
parametric and non-parametric decision models, geometric patterns,
shape similarity measures.
Neural networks are not treated to avoid overlap with the course Neural Networks,
logic and reasoning aspects are not treated to avoid overlap with the
course Probabilistic Reasoning.