Literature Review of Meta-Heuristic Methods

Literature Review of Meta-heuristic Methods

Name

Institution

Author note

We can define meta-heuristic as a high-level algorithmic framework of problem independent that gives a set of procedures and strategies for the development optimum heuristic algorithms. There are several examples of meta-heuristics, which include genetic or evolutionary algorithm, ant colony optimization, simulated annealing, and tabu search among many other forms. According to the available guideline, a specific-problem implementation of optimization algorithm heuristic expressed in the framework of meta-heuristic is also called a meta-heuristic. The algorithms of meta-heuristic, which means optimization procedures designed according the strategies in the framework of meta-heuristic are always heuristic in nature (Bianchi et al. 2009). This fact differentiates them from the precise methods that come with a proof to indicate that optimal solutions will be in the form of finite amount of time. Therefore, meta-heuristic are specifically built in order to find solutions that are appropriate for the computation of small amount of time. Consequently, they are not subject to combinatorial explosion, which is a computing phenomena that needs computing time in order to determine the necessary solutions. In this case, hard problems increase in an exponential with respect to the size of the problem. The aim of this paper is to provide a literature review of the article by Kenneth & Fred (2009) about their survey on meta-heuristics.

The scientific communities have been demonstrating Meta-heuristics as the most viable and more superior alternative as compared to the conventional or exact procedures of the mix-integer optimization. These include branch and bound and dynamic programming procedures, especially for complex problems or large instances of problems (Bianchi et al. 2009). Often, the meta-heuristics have the ability to provide a better distinction between the quality of solution and the time of computing. In addition, meta-heuristics are more flexible in their application as compared to the conventional or exact methods in several ways. To begin with, because the framework of meta-heuristics is having general definition, we can adapt the algorithms of meta-heuristic in order to fit the requirements of real life problems of optimization with regard to the stipulated quality of solutions and allowed time for computation. However, this can have great variations across various problems and situations.

Secondly, meta-heuristics do not place any formulation demands for optimization problems such as constraints requirements as a linear function of the variable used for making decisions. Nevertheless, this aspect of flexibility comes at a considerable cost that requires adaptation of specific problem in order to attain the best performance. This is the reason why the meta-heuristics research field involves many critics (Danna & Le 2005). Many of the critics base their arguments on the fact that meta-heuristics procedures lack universal application designs for their methodological approaches to problems. They argue that the field of meta-heuristics lack scientific severity when it comes to testing various implementations. They also argue that the field has a tendency of building complex procedures and applying different operators that many people without the field's knowledge cannot comprehend (Danna & Le 2005). However, many of these arguments have failed to produce a successful story without concrete examples.

One of the major reasons why meta-heuristics is the best method of choice for solving many large-scale real-life problems is that, it has the ability to provide better solutions as than other exact procedures. This is true both in academic research works and in practical application situations. Because of this, many vendors of commercial software have implemented meta-heuristic techniques as their fundamental engines for optimization (Bianchi et al. 2009). They have applied these techniques both in specialized packages of software for scheduling production, routing of vehicles, planning duty rosters for nurses among other complex problems. These also include optimization for general purpose and software simulation techniques. The underlying foundation of various procedures of meta-heuristics varies significantly. Some of the methods model the process of optimization by using a metaphor that is not related to optimization (Bianchi et al. 2009). The best example in this case is the natural evolution or the evolutionary algorithms, cooling of solid crystals using simulated annealing, and determination of animals' behaviors while in swarms such the optimization of ant colony.

However, other meta-heuristics methods such as the tabu search do not apply such intermediary levels of description. Instead, they focus on the structure of problem exploitation in order to improve their search of achieving the best results. Generally, frameworks of meta-heuristics rely heavily on the use of random applications, even though some researchers have proposed several deterministic strategies. Many frameworks of meta-heuristics originated in 1980s; even though we can trace some cases of meta-heuristic methods back in 1960s and 1970s. The methods have enjoyed steady growth in their both applications and popularity since 1980s. The field of meta-heuristics is currently the subject that several journals and conferences have dedicated their time to discuss. For example, EU/ME is a meta-heuristics community-working group whose sponsor is the EURO (Danna & Le 2005). With approximately 1400 members, this group is the largest among the researchers in the field of meta-heuristics throughout the world.

Taxonomy of Meta-heuristics

Taxonomy of meta-heuristic is one of the major topics that come out clearly from the article. The meta-heuristic algorithms try to come up with the best financial solutions in all their solutions of every optimization problem (Bianchi et al. 2009). At this point, the algorithms evaluate potential solutions to optimization problems and perform a series of operations on them in order to find various solutions and determine the best solutions. Meta-heuristics procedures operate on a solution representation or on a solution encoding. This feature of meta-heuristics can be stored in the memory of the computer. In addition, the user can conveniently manipulate these procedures using different meta-heuristic operators. We can distinguish the three classes of meta-heuristics based on the manner in which we manipulate solutions. For instant, the local search meta-heuristics make iterative small changes in