Details
- Supervisors
- Faculty
- Degree label
- Abstract
- The problem of minimizing an objective function f on a Riemannian manifold has been a topic of much interest over the past few years due to many important applications, such as low-rank learning, shape analysis, image segmentation, matrix mean computation, and blind source separation. Several Riemannian optimization algorithms already exist that converge to local minima of f, but the research on Riemannian global optimization algorithms is still limited. In contrast, global optimization on Euclidean spaces has been investigated for a long time and many algorithms have been proposed, e.g., Particle Swarm Optimization, Bacterial Foraging Optimization, Random Linkage method, Markov Chain Monte Carlo (MCMC) Simulated Annealing and Covariance Matrix Adaptation Evolution Strategy. The purpose of this master's thesis is to investigate a few of these global optimization algorithms that are relevant on Euclidean spaces and then generalize at least two of them to the manifold setting. The generalization should address the algorithm descriptions and also their analysis. A package needs to be developed and tested on one or more applications. In this work, I investigated the performances of two new methods with their accelerated versions and compare their performances with the already implemented ones on the manifold setting.