National Technical University of Athens (NTUA)
Καινοτόμες μεθοδολογίες μηχανικής μαθήσεως
Abstract
dc:descriptionThis thesis aimed at the extensive study of modem machine learning methodologies, and the provision of novel methodologies and groundbreaking regards towards pattern recognition techniques. A great deal of emphasis was put on Variational Bayesian inference techniques, which, to the author's mind, comprise the future of statistical clustering-based approaches, resulting in the provision of a novel model for robust recognition of high-dimensional patterns. Another key concept studied was fuzzy clustering. In this latter field can be found the most significant and seminal contribution of this thesis: we have introduced a novel regard towards fuzzy clustering, by showing that fuzzy clustering techniques, in the form of FCM algorithm variants, provide an interesting alternative to the EM algorithm (and other statistical clustering alternatives) for the treatment of probabilistic generative models. Furthermore, this work provided a novel hidden Markov model, offering significant benefits comparing to modem sequential data modeling techniques, as well as a new speaker verification methodology, based on Gaussian process classifiers.
Degree
thesis:*- Grantor dc:publisher
- National Technical University of Athens (NTUA)
- Year dc:date
- 2008
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Chatzis, Sotirios
- Χατζής, Σωτήριος
Subjects
dc:subject × 14- Μηχανική μάθηση
- Αναγνώριση προτύπων
- Πιθανοτικά παραγωγικά μοντέλα
- Κατάτμηση εικόνων
- Μαρκοβιανά μοντέλα
- Machine learning
- Pattern recognition
- Probabilistic generative models
- Image segmentation
- Markovian models
- 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
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/17396