Browsing by Subject "PSO"

Now showing 1 - 6 of 6

Access status: Restricted ,
Agentowa implementacja algorytmu PSO
(Data obrony: 2019-09-27) Nyznar, Mariusz
Wydział Elektrotechniki, Automatyki, Informatyki i Inżynierii Biomedycznej
Access status: Restricted ,
Computational intelligence algorithms for vehicle path planning problem
(Data obrony: 2020-07-09) Pudełko, Radosław
Wydział Elektrotechniki, Automatyki, Informatyki i Inżynierii Biomedycznej
Access status: Restricted ,
Dobór nastaw regulatora za pomocą algorytmu ptasiego
(Data obrony: 2018-01-18) Szczerba, Joanna
Wydział Elektrotechniki, Automatyki, Informatyki i Inżynierii Biomedycznej
Access status: Restricted ,
Implementacja algorytmu populacyjnego z pamięcią długo- i krótkoterminową.
(Data obrony: 2019-10-25) Mierzwa, Maria
Wydział Elektrotechniki, Automatyki, Informatyki i Inżynierii Biomedycznej
Access status: Restricted ,
Implementacja metody roju cząstek w OpenCL
(Data obrony: 2014-12-11) Iwaniak, Michał
Wydział Geologii, Geofizyki i Ochrony Środowiska
Presented thesis describes Particle Swarm Optimization algorithm, a type of a heuristic algorithm used for many optimization problems. The objective of this paper was to implement PSO utilizing computing power of graphics processing units (GPU) emphasizing research of the impact of selected configuration parameters. Cross-platform availability was the main reason why OpenCL was chosen as the framework in which the algorithm was implemented. Main bottlenecks of GPU parallel implementation were high global memory access time and high execution time of random number generator functions. The former was eliminated by removal of the critical section from the algorithm, while the latter by implementing own RNG algorithm, that executed directly on the kernel.
Access status: Open Access ,
On obtaining effective elasticity tensors with entries zeroing method
(Wydawnictwa AGH, 2018) Gierlach, Bartosz; Danek, Tomasz
The purpose of this paper is to propose a new method for obtaining tensors expressing certain symmetries, called effective elasticity tensors, and their optimal orientation. The generally anisotropic tensor being the result of <i>in situ</i> seismic measurements describes the elastic properties of a medium. It can be approximated with a tensor of a specific symmetry class. With a known symmetry class and orientation, one can better describe geological structure elements like layers and fissures. A method used to obtain effective tensor in the previous papers (i.e. Danek & Slawinski 2015) is based on minimizing the Frobenius norm between the measured and effective tensor of a chosen symmetry class in the same coordinate system. In this paper, we propose a new approach for obtaining the effective tensor with the assumption of a certain symmetry class. The entry zeroing method assumes the minimization of the target function, being the measure of similarity with the form of the effective tensor for the specific class. The optimization of orientation is made by means of the Particle Swarm Optimization (PSO) algorithm and transformations were parameterised with quaternions. To analyse the obtained results, the Monte-Carlo method was used. After thousands of runs of PSO optimization, values of quaternion parts and tensor entries were obtained. Then, thousands of realizations of generally anisotropic tensors described with normal distributions of entries were generated. Each of these tensors was the subject of separate PSO optimization, and the distributions of rotated tensor entries were obtained. The results obtained were compared with solutions of the method based on the Frobenius distances (Danek et al. 2013).