Level
KI.B2
|
Prerequisites
At least a full year of prior study in the field of cognitive science, artificial intelligence, or computer science. Furthermore, (2nd year) experience in programming is required.
|
Contents
The course BKI230 provides students with a thorough understanding of artificial neural network (ANN) models, by exploring fundamental properties like architectural aspects and the processes of information processing, learning, and classification. Among the models covered are feedforward nets (perceptrons, backprop), competitive learning (Grossberg, Kohonen), constraint satisfaction (Hopfield), and modular architectures and recurrent architectures. Furthermore, students will learn more about the practical application of, e.g., multi-layered perceptrons and Kohonen neural networks. The complete pattern recognition pipeline (data acquisition, pre-processing, segmentation, feature extraction, and classification) will be covered. Central theme of the pattern recognition process reads: "How can low-level, sub-symbolic features be transformed into higher-level symbolic features?" In the process of the recognition of handwritten shapes (handwriting, geometrical objects), students will learn about signal processing, feature extraction, data analysis and clustering, and training and testing neural network simulations. Students are required to program feature extraction algorithms and they will participate in a challenge "Who achieves the highest recognition results"! The results of this exploration in feature and classifier space must be documented in an end report, evaluating your findings
|
Literature
See website for more details (http://www.ai.ru.nl/aicourses/bki230.
|
Teaching methods
Lectures and practical hours.
|
Information
dr. L.G. Vuurpijl, B.02.33, tel. 3615981, e-mail:
|
Coördinerend docent
dr. L.G. Vuurpijl
|
Exam
practicumopdrachten en 2 deeltentamens (tijdens college-uren, halverwege en aan het eind v/h semester)
|
Exam information
Exams. practical exercises and end report.
|
Enrollment
verplicht via KISS/TIS tot 2 werkdagen voor aanvang; deze inschrijving geldt tevens voor het tentamen |
Extra information
Language of instruction Dutch or English
|
