Evolutionary computing technique

Despite a mass of research activity in evolutionary computation (EC), activity that has led to solid theoretical results and realistic applications, there are still a number of perennial irritations that almost all EC techniques suffer from. First, the serious computational cost of evaluating numerous potential solutions (or individuals), over hundreds of iterations (or generations), places pragmatic and sometimes formal limitations on the use of EC in real-world applications with time-sensitive outputs, such as online multiprocessor scheduling. This real limitation deters many potential users from using or even considering the use of EC in heavy-duty engineering and scientific applications. 

The authors present some of the methods that we have developed and exploited in Aberystwyth for gathering highly multivariate data from bioprocesses, and some techniques of sound multivariate statistical analyses (and of related methods based on neural and evolutionary computing) which can ensure that the results will stand up to the most rigorous scrutiny. 

The boundaries to AI, within which evolutionary optimisers He, are woolly, and this may make it appear a mysterious subject to those who are unfamiliar with it even those whose research lies in AI find it hard to agree on the precise boundaries to the field. It encompasses a large family of algorithms, from neural networks and knowledge-based systems to those within evolutionary computing. Almost every major AI technique is now used within chemistry (Table 1), and in an increasing number of cases, AI is the method of choice. 

Although the use of evolutionary computing within science is still in its early stages, scientific applications are already notable for their diversity. Evolutionary methods have been used to optimise the geometry of molecules, the shape of propellers, the properties of polymers and the order in which chemicals are produced in industrial flow fines. They have been used in the study of oil extraction, the natural degradation of toxic chemicals in the environment, spectral deconvolution, the interpretation of microwave spectra and in a wide range of other areas. The chapters that follow provide further illustration of the potential of these intriguing and versatile techniques. 

AI techniques can be roughly divided into two categories symbolic AI and computational intelhgence. The former focuses on development of knowledge-based systems while the latter focuses on development of a set of nature-inspired computational approaches. The latter primarily includes evolutionary computations, artificial nemal networks and fuzzy logic systems. A brief introduction to these techniques begins on the next page. 

Evolutionary computation is an umbrella term for a range of evolutionary optimization techniques mainly inspired by optimum-seeking mechanisms from the real world, such as natural selection and genetic inheritance, which simulate evolution processes on a computer to iteratively improve the performance of solutions until an optimal (or feasible at least) solution is obtained. 

The processes of optimum-seeking have been remarkably successful in lots of real-world phenomena, such as human evolution, food-seeking of ant colonies, and improvisation of musicians. By using stochastic heuristic individual searches and generation processes, these phenomena work toward a perfect individual to fill a particular environmental niche. It is naturally expected that evolutionary optimization processes can be created by modeling the behaviors of these phenomena. The evolutionary optimization techniques were thus developed to perform this function, which mimics the optimum-seeking processes of these phenomena in a computer program. 

Particle Swarm Optimization (PSO) is a stochastic optimization method evolved from Swarm Theory and Evolutionary Computation [4]. It is instigated by animals natural swarming behavior [5]. PSO has been proven to be a suitable technique for solving various optimization problems [6, 7]. Among the advantages of PSO are that it allows efficient and rapid optimization of the problem, due to its parallel nature, it requires only basic mathematical operators for optimization and it provides low computational and memory costs for each iteration [8]. Many variants of the PSO algorithm exist, such as PSO with inertia weight [9], PSO with constriction factor [10], and mutative PSO [11]. 

Population-based methods, also known as evolutionary computations, search the entire solution space S by maintaining a group of candidate vectors. Evolutionary techniques are inspired by natural evolution and adaptation with the essence of survival of the fittest. During the iterative process, new candidate vectors are... 

AI methods may be used in various ways. The models may be used as a standalone application, e.g., in recent work on the design of microwave absorbers using particle swarm optimization (PSO).6 Alternatively, a computational tool, such as a finite element analysis or a quantum mechanical calculation, may be combined with an AI technique, such as an evolutionary algorithm. 

Coello Coello, C.A. (2002) Theoretical and numerical constraint handling techniques used with evolutionary algorithms A survey of the state of the art. Computer Methods in Applied Mechanics and Engineering, 191 (11—12), 1245-1287. 

The methods of simulated annealing (26), genetic algorithms (27), and taboo search (29) are three of the most popular stochastic optimization techniques, inspired by ideas from statistical mechanics, theory of evolutionary biology, and operations research, respectively. They are applicable to our current problem and have been used by researchers for computational library design. Because SA is employed in this chapter, a more-detailed description of the (generalized) SA is given below. 

Genetic programming [137] is an evolutionary technique which uses the concepts of Darwinian selection to generate and optimise a desired computational function or mathematical expression. It has been comprehensively studied theoretically over the past few years, but applications to real laboratory data as a practical modelling tool are still rather rare. Unlike many simpler modelling methods, GP model variations that require the interaction of several measured nonlinear variables, rather than requiring that these variables be orthogonal. 

---