|
|
Genetické algoritmy
Masárová, Mária ; Zbořil, František (oponent) ; Zbořil, František (vedoucí práce)
Táto práca sa zaoberá genetickými algoritmami, ich terminológiou a využitím. Popisuje rôzne problémy, ktoré sa dajú pomocou genetických algoritmov riešiť. V práci sú taktiež predstavené rôzne algoritmy skupinovej inteligencie, pričom algoritmus svetlušiek slúži aj na porovnanie efektivity medzi ním a genetickým algoritmom. Hlavnou úlohou tejto práce je vykonať experimenty s tromi optimalizačnými úlohami, konkrétne sú to, problém obchodného cestujúceho, splniteľnosť logických formúl a hľadanie extrému funkcie.
|
|
|
Evolutionary Design of Ultrasound Treatment Plans
Masárová, Mária ; Sekanina, Lukáš (oponent) ; Jaroš, Jiří (vedoucí práce)
The use of focused ultrasound helps save and facilitate human lives, as it can be used to treat epilepsy, destroy cancer cells and stop internal bleeding in a non-invasive way, which is a more acceptable and safer solution for humans. Given that human safety and health is a priority in the treatment of serious diseases, this work deals with the comparison of different evolutionary algorithms and their use in the design of evolutionary ultrasound treatment plans. Two types of media are used in this work, namely homogeneous medium and heterogeneous medium. When evaluating algorithms, we focus on efficiency with respect to the size of the population, number of fitness function evaluations and computational time. In a homogeneous medium, CMA-ES proved to be the best algorithm, which was able to find the optimal solution with a 100% coverage of the target area for a rotated grain of rice within 20 seconds. However, the heterogeneous medium is a much more complex problem, mainly due to the skull, which reflects and absorbs much of the ultrasound. Here, the SA algorithm proved to be the best, finding the result with a 23% coverage of the target area in the first test scenario. The calculation time took about 1 hour and 18 minutes, which means that time is a very prohibitive factor.
|
|
|
Evolutionary Design of Ultrasound Treatment Plans
Masárová, Mária ; Sekanina, Lukáš (oponent) ; Jaroš, Jiří (vedoucí práce)
The use of focused ultrasound helps save and facilitate human lives, as it can be used to treat epilepsy, destroy cancer cells and stop internal bleeding in a non-invasive way, which is a more acceptable and safer solution for humans. Given that human safety and health is a priority in the treatment of serious diseases, this work deals with the comparison of different evolutionary algorithms and their use in the design of evolutionary ultrasound treatment plans. Two types of media are used in this work, namely homogeneous medium and heterogeneous medium. When evaluating algorithms, we focus on efficiency with respect to the size of the population, number of fitness function evaluations and computational time. In a homogeneous medium, CMA-ES proved to be the best algorithm, which was able to find the optimal solution with a 100% coverage of the target area for a rotated grain of rice within 20 seconds. However, the heterogeneous medium is a much more complex problem, mainly due to the skull, which reflects and absorbs much of the ultrasound. Here, the SA algorithm proved to be the best, finding the result with a 23% coverage of the target area in the first test scenario. The calculation time took about 1 hour and 18 minutes, which means that time is a very prohibitive factor.
|
|
|
Genetické algoritmy
Masárová, Mária ; Zbořil, František (oponent) ; Zbořil, František (vedoucí práce)
Táto práca sa zaoberá genetickými algoritmami, ich terminológiou a využitím. Popisuje rôzne problémy, ktoré sa dajú pomocou genetických algoritmov riešiť. V práci sú taktiež predstavené rôzne algoritmy skupinovej inteligencie, pričom algoritmus svetlušiek slúži aj na porovnanie efektivity medzi ním a genetickým algoritmom. Hlavnou úlohou tejto práce je vykonať experimenty s tromi optimalizačnými úlohami, konkrétne sú to, problém obchodného cestujúceho, splniteľnosť logických formúl a hľadanie extrému funkcie.
|
host ::
RSS.