作者: Godfrey C Onwubolu , BV Babu , Pablo Moscato , Alexandre Mendes , Alexandre Linhares
DOI: 10.1007/978-3-540-39930-8_18
关键词:
摘要: With applications ranging from fields as distinct fuzzy modeling (Xiong 2001), autonomous robot behavior (Luk et al. learning with backpropagation (Foo 1999), and multicriteria optimization (Viennet 1996), evolutionary methods have become an indispensable tool for systems scientists. Although already studied in the past, interesting emerging issue is use of multiple populations, which gaining increased momentum conjunction two technologies. On hardware side, computer networks, multi-processor computers distributed processing (such workstations clusters) are increasingly becoming widespread. Regarding software issues, introduction Parallel Virtual Machine1 (PVM), later Message Passing Interface Standard2 (MPI), well web-enabled, object-oriented languages Java3) also had their role. Most Evolutionary Algorithms (EAs) that easy to parallelize naturally suitable heterogeneous systems. For most EAs distribution tasks relatively applications. The workload can be at individual or a population level; final choice depending on complexity computations involved.