Technical University of Crete (TUC)
RECURRENT HIGH ORDER NEURAL NETWORKS FOR LEARNING IN DYNAMIC ENVIRONMENTS
Abstract
dc:descriptionIN THIS THESIS, THE APPLICABILITY OF RECURRENT HIGH ORDER NEURAL NETWORKS (RHONNS) WHEN USED IN IDENTIFICATION OF UNKNOWN NONLINEAR DYNAMICAL SYSTEMS IS EXAMINED. RHONNS ARE SHOWN TO BE CAPABLE OF APPROXIMATING-TO ANY DEGREE OF ACCURACY- A QUITE GENERAL CLASS OF NONLINEAR SYSTEMS. APPROPRIATE LEARNING LAWS-WHICH ARE BASED ON WELL KNOWN ADAPTIVE ROBUST PARAMETER ESTIMATION TECHNIQUES-ARECONSTRUCTED AND IT IS SHOWN THAT THESE LEARNING LAWS ARE GLOBALLY CONVERGENT,STABLE AND ROBUST; USING THESE LEARNING LAWS. THE IDENTIFICATION ERROR IS SHOWN TO CONVERGE TO A BALL CENTERED AT THE ORIGIN, AND WHOSE RADIUS CAN BE MADE ARBITRARILY SMALL BY INCREASING THE NUMBER OF RHONN HIGH ORDER CONNECTIONS. A TOTALLY NOVEL LEARNING LAW IS PROPOSED WHICH IS NOT ONLY CONVERGENT, STABLE AND ROBUST, BUT IS ALSO ENSURES EXPONENTIAL IDENTIFICATION ERROR CONVERGENCE FOR ANY RHONN ARCHITECTURE. IN OTHER WORDS, THE CONVERGENCE OF THE NEW LEARNING LAW IS INDEPENDENT OF THE NUMBER OF RHONN HIGH ORDER TERMS, AND, MOREOVER, SUCH A CONVERGENCE IS THE BEST THAT CAN BE ACHIEVED. AT LAST, WE MENTION THAT THERATE OF CONVERGENCE DEPENDS ON A DESIGN PARAMETER Γ, AND WE CAN MAKE THE RATEOF CONVERGENCE ARBITRARILY LARGE BY SIMPLY INCREASING Γ.
Degree
thesis:*- Grantor dc:publisher
- Technical University of Crete (TUC)
- Year dc:date
- 1995
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Kosmatopoulos, Elias
- Κοσματόπουλος, Ηλίας
Subjects
dc:subject × 20- ΑΛΓΟΡΙΘΜΟΙ ΕΚΜΑΘΗΣΗΣ ΚΑΙ ΠΡΟΣΑΡΜΟΓΗΣ
- ΔΙΑΝΥΣΜΑΤΙΚΗ ΚΒΑΝΤΟΠΟΙΗΣΗ
- ΕΥΣΤΑΘΕΙΑ ΚΑΤΑ LYAPUNOV
- ΜΗ-ΓΡΑΜΜΙΚΑ ΣΥΣΤΗΜΑΤΑ
- Δίκτυα, Νευρωνικά
- Ρομποτική
- ΣΥΓΚΛΙΣΗ ΑΛΓΟΡΙΘΜΩΝ
- ΤΑΥΤΟΠΟΙΗΣΗ ΣΥΣΤΗΜΑΤΩΝ
- ALGORITHM CONVERGENCE
- LEARNING AND ADAPTIVE ALGORITHMS
- Lyapunov stability
- Neuronic networks
- NON-LINEAR SYSTEMS
- Robotics
- System identification
- VECTOR QUANTIZATION
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Engineering and Technology
- Electrical Engineering, Electronic Engineering, Information Engineering
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/3764