Abstract
dc:descriptionStatistical methods for signal and image processing utilize probabilistic models for the data sets and sampling mechanisms. The statistical properties of the data and sampling mechanisms typically correspond to special geometric relations. This thesis examines the problems of phase retrieval as well as the training of fully connected neural networksin the context of image classification. Both are image processing problems equivalent to non linear and non convex mathematical optimization. The theoretical and experimental investigation of the statistical properties present in the mathematical modelsof the aforementioned problems, as well as their corresponding implicit geometric relations, inspired the design of novel initialization and optimization methods. The resulting algorithms led to significantly lower execution times for comparable or superior solution quality, compared to state of the art methods.
Degree
thesis:*- Grantor dc:publisher
- National and Kapodistrian University of Athens
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Nyfantis, Petros
- Νυφαντής, Πέτρος
Subjects
dc:subject × 26- Στατιστικές
- Μέθοδοι
- Επεξεργασία
- Εικόνα
- Ανάκτηση
- Φάσης
- Νευρωνικά
- Δίκτυα
- Εκπαίδευση
- Αρχικοποίηση
- Statistical
- Methods
- Image
- Processing
- PHASE
- Retrieval
- Neural
- Networks
- Training
- Initialization
- Άλλες Επιστήμες Μηχανικού και Τεχνολογίες
- Επιστήμες Μηχανικού και Τεχνολογία
- Τεχνολογία απεικόνισης και Φωτογραφική τεχνολογία
- Other Engineering and Technologies
- Engineering and Technology
- Imaging Science and Photographic Technology
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/59775