Novel parallelization of simulated annealing and Hooke & Jeeves search algorithms for multicore systems with application to complex fisheries stock assessment models

详细信息    查看全文

• 作者：Sergio Vá ; zquez^a ; ^{sergio.vazquez@udc.es}Author Vitae ; Mar&iacute ; a J. Mart&iacute ; n^a ; ^{mariam@udc.es}Author Vitae ; Basilio B. Fraguela^a ; ^{basilio.fraguela@udc.es}Author Vitae ; André ; s Gó ; mez^b ; ^{agomez@cesga.es}Author Vitae ; Aurelio Rodr&iacute ; guez^b ; ^{aurelio@cesga.es}Author Vitae ; Bjarki Þ ; ó ; r Elvarsson^c ; ^{bthe@hafro.is}Author Vitae
• 关键词：Marine ecosystems ; Gadget ; Simulated annealing ; Hooke & Jeeves ; Multicore systems ; OpenMP
• 刊名：Journal of Computational Science
• 出版年：2016
• 出版时间：November 2016
• 年：2016
• 卷：17
• 期：part_P3
• 页码：599-608
• 全文大小：2521 K
• 卷排序：17

文摘

The most costly part of an ecosystem model is the optimization of its parameters. This involves search strategies such as Simulated Annealing or Hooke & Jeeves. We propose two novel OpenMP-based parallel versions of each one of these algorithms. Our proposals are totally general and do not reduce the quality of the solution. They allow to make a better use of today's mainstream multicore processors.