Real time learning in Intelligent Systems)

Recommended Prerequisites

MSc in Informatics Engineering or equivalent. Knowledge about neural networks, fuzzy logic and discrete systems.

Teaching Methods

Theoretical classes (2h per week) for presentation and discussion of the material of each module and for presentation of the synthesis and research works by the students (one presentation per module).

Adopted resources:
- slides of theoretical pectures
- notes (text) about some parts of the syllabus
- miscellaneous bibliography
- recent papers in journals
- software: Matlab, Simulink and Toolboxes.

Learning Outcomes

Competency to develop computational models in real time from data following two paradigms: quantitative data and qualitative data. The parameters of these models are trained on-line, in real time using artificial neural networks and support vector machines for the quantitative paradigm and fuzzy systems for the qualitative paradigm.

Generic competencies in analysis and synthesis, written and oral communication, informatics knowledge relative to the study focus, problem solving, critical thinking, decision capability, autonomous learning, research, practical application of theoretical knowledge, creativity, self-criticism and self-evaluation, research.

Work Placement(s)

No

Syllabus

1.    Linear recursive models from data: recursive parameter identification for ARX,  ARMAX and other discrete time series models.
2.    Real-time training of recurrent neural networks
3.    Recursive training of Support Vector Machines
4.    Evolving Fuzzy Systems: real time building of fuzzy rules, online updating of the rule base. Accuracy,  transparency and interpretability.

Head Lecturer(s)

António Dourado Pereira Correia

Assessment Methods

Assessment
Synthesis work: 50.0%
Research work: 50.0%

Bibliography

Systems Identification, theory for the user,  Leonard Ljung:, Prentice Hall; 2 edition (January 8, 1999)
Evolving Fuzzy Systems, Angelov P, D Filev, N Kasabov Eds., Evolving Intelligent Systems: Methodology and Applications, John Willey and Sons, 2008,
Evolving Fuzzy Systems - Methodologies, Advanced Concepts and Applications Series: Studies in Fuzziness and Soft Computing, Vol. 266  Lughofer, Edwin, Springer Verlag.
Evolving fuzzy systems for data streams: a survey, Rashmi Dutta Baruah, Plamen Angelov* Article first published online: 28 JUL 2011 DOI: 10.1002/widm.42 Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery Volume 1, Issue 6, pages 461–476, November/December 2011, John Wiley & Sons, Inc.
Bernhard Schölkopf and Alex Smola. Learning with Kernels. MIT Press, Cambridge, MA, 2002.Papers from international journals