Institutes outside Greece
Randomized derivative free optimization via CMA-ES and sparse techniques: applications to radars
Abstract
dc:descriptionIn this thesis, we investigate aspects of adaptive randomized methods for black-box continuous optimization. The algorithms that we study are based on the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) algorithm and focus on large scale optimization problems. We start with a description of CMA-ES and its relation to the Information Geometric Optimization (IGO) framework, succeeded by a comparative study of large scale variants of CMA-ES. We furthermore propose novel methods which integrate tools of high dimensional estimation within CMA-ES, to obtain more efficient algorithms for large scale partially separable problems. Additionally, we describe the methodology for algorithm performance evaluation adopted by the Comparing Continuous Optimizers (COCO) platform, and finalize the bbob-largescale test suite, a novel benchmarking suite with problems of increased dimensions and with a low computational cost. Finally, we present the formulation, methodology and obtained results for two applications related to Radar problems, the Phase Code optimization problem and the Phased-Array Pattern design problem.
Degree
thesis:*- Grantor dc:publisher
- Institutes outside Greece
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Varelas, Konstantinos
- Βαρελάς, Κωνσταντίνος
Subjects
dc:subject × 14- Αλγόριθμοι βελτιστοποίησης
- Numerical optimization
- Μαθηματικά
- Φυσικές Επιστήμες
- Έλεγχος και Βελτιστοποίηση
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Υπολογιστές, Υλικό (hardware) και Αρχιτεκτονική
- Mathematics
- Natural Sciences
- Control and Optimization
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Computer science, Hardware and Architecture
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/55697