英语论文网

ther than deterministic algorithms. Deterministic algorithms guarantee to find an optimal solution in bounded time for every finite size instance of a problem, but they often lead to computation times too high for practical purposes. As approximate algorithms, metaheuristic sacrifice the guarantee of finding optimal solutions for the sake of getting good solutions in a significantly reduced amount of time (Ong, Lim, Zhu & Wong, 2006). Many of the metaheuristic approaches rely on probabilistic decisions made during the search. But, the main difference to pure random search is that in metaheuristic algorithms randomness is not used blindly but in an intelligent, biased form. Some widely adopted meta-heuristic methods include but not limited to Simulated Annealing (SA) (Osman, 1993), Tabu Search (TS) (Osman, 1993), Genetic Algorithms(GA) (Wehn, 1991) and Particle Swarm Optimization (Kennedy & Eberhart, 2001).

1.3.2 对共通启发法的分类——1.3.2 Classification of Metaheuristics

This thesis is concerned with metaheuristic algorithms, which are general methods for solving search and optimization problems. Although these algorithms exist today in many different forms, they all search for solutions in essentially the same way. That is, they try out different alternatives, evaluate their worth and then repeat this procedure using what has been learnt to guide them. What differentiates the numerous algorithms, then, is the way that they utilize the information from the solutions they have explored in the past, and that they have in some sense “remembered”, in order to find or construct better solutions in the future.

There are several possible classifications of heuristics and metaheuristics but one that is commonly used and that certainly allows us to embrace most metaheuristics including their hybrids is: single agent search where a single solution is improved incrementally by making small changes to it; and population-based methods where a population of solutions are ‘evolved' in parallel, and improvement results from repeatedly making variations of those solutions in the population that are more highly ‘fit', and discarding those that are less fit (Blum & Roli, 2001).

Examples of single-solution methods are: basic local search (deterministic iterative improvement), simulated annealing, tabu search, greedy randomized adaptive search procedure, variable neighbourhood search, guided local search, iterated local search and others. Population-based methods include: genetic algorithms, scatter search, ant colony systems, memetic algorithms, evolutionary strategies (although some of them are single-agent search), particle swarm systems, cultural algorithms, etc. If a single-agent search approach is hybridised with a population based approach (e.g. a memetic algorithm can be defined to be a genetic algorithm incorporating local search) then the result is, of course, a population-based approach.

Sometimes, researchers classify heuristic and metaheuristic approaches into nature-inspired and non-nature inspired and many refer to the first group as evolutionary algorithms. While these algorithms are commonly conceptualized as those approaches that simulate various aspects of natural evolution (B?ck, Fogel & Michalewicz, 1997), some researchers argue that a fundamental characteristic of evolutionary algorithms is that they handle a 本论文由英语论文网提供整理，提供论文代写，英语论文代写，代写论文，代写英语论文，代写留学生论文，代写英文论文，留学生论文代写相关核心关键词搜索。