University of Twente Student Theses


Recognition of Structures in Numerical Data

Iakovou, Dimitrios (2003) Recognition of Structures in Numerical Data.

[img] PDF
Abstract:In control engineering, system identification is of the utmost importance for determining the correct control parameters for a plant. In the present research we investigate the structures present in numerical plant data and derive symbolic expressions that correspond to these structures. Therefore, this research can also be applied to the inverse problem. For the development of such a recognition system, we will make use of the morphogenetic neuron. Like classical neuron networks, the Morphogenetic neurons are capable in recognizing, but in a higher level of abstraction. The Morphogenetic neurons are able to encode abstract, symbolic expressions that characterize the relations between the inputs and outputs of a system. Generally, in most measured data, there are (hidden) underlying relations (correlations, invariants, rules). Our approach is to automatically recognize these underlying structures by observing the dimensionality of the basis that is required for representation of the data. The basis is approximated by a morphogenetic neuron. The basic problem is the selection of the number and the type of the variables that will participate in the basis functions. Additionally, it is desired not to make any a-priori assumptions on the properties of the numerical data and let the Morphogenetic neuron derive to a solution unaided. Within this project, the research has been limited to 2-dimensional space for computational reasons, but this technique is generally applicable to n-dimensional space.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:53 electrotechnology
Programme:Electrical Engineering MSc (60353)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page