并行进化算法中适应迁移间隔的方案设计与分析.pdfVIP

Evolutionary Computation corrected proof doi:10.1162/EVCOa00153 o by the Massachusetts Institute of Technology Evolutionary Algorithms Andrea Mambrini Dirk Sudholt Abstract with regard to running time and communication effort. Keywords theory. 1 Introduction for tackling large-scale problems. o by the Massachusetts Institute of Technology Evolutionary Computation corrected proof doi:10.1162/EVCOa 00153 eby the Massk astekar regions of the search space. 2011; Alba et al., 2013). of the current migration interval. 2 Eoolutionary Computation corrected proof doi:10.1162/EVCOa 00153 problem classes, we can only compare upper bounds. 1.1 Related Work problem. optimization problems. stops prematurely. (2006). Evolutionary Computation Volume xx, Number x 3 Evolutionary Computation corrected proof doi:10.1162/EVCOa00153 eby the Massk a stakar fixed-length runs. interval are inspired by this work. 2 Preliminaries after ti generations, the migration interval t, doubles. 4 Evolutionary Computation corrected proof doi:10.1162/EVCO a 00153 all1iλ 3: fort;=1 to x do 4: for all1≤i≤λ in parallel do 7: if u;≥πi or ftf; then 8: 9: end if 11: ifu;≥πiorfifi then 12: if ffi then 13: Letti:=1 and ft:=fi 14: else 15: Letπi:=Ti·2 16: end if 17; Let u;;=0 18: end if 19:Mj:= +1 20: end for 21: end for communication effort has to be multiplied by v. special case where each island ru