Kunjal Bharatkumar Mankad | IJCSET(www.ijcset.net) | August 2015 | Vol 5, Issue 8,304-308 A Comparative Analysis of Design Issues of Evolutionary Structures Kunjal Bharatkumar Mankad Rajkot, India Abstract-Evolutionary computing is ubiquitous in nature. comparison among evolutionary characteristics such as Adaptive nature and evolutionary search capabilities of representation, encoding, and types of genetic operators. evolutionary structures produce a more efficient exploration The concluding part of the paper justifies the design of the state space of possible solutions. Evolutionary perspective of evolutionary structures. computing provides four major structures namely evolutionary programming, genetic programming, genetic algorithm and evolutionary strategy. The uniqueness of II. EVOLUTIONARY ALGORITHMS evolutionary structure is that they are capable to provide In recent years, cognitive systems have gained prominence solutions for highly complex mathematical problems very by implementing evolutionary approach to the efficiently. The paper explains differences between traditional computational modeling. Computational paradigm which is search and optimization algorithm and evolutionary algorithm based on evolutionary characteristics is popularly known as along with advantages of evolutionary computing. evolutionary computing. Evolutionary computing is based Evolutionary life cycle and procedural characteristics of each on principles of natural genetics. It is basically the type of structure are discussed in details. Structural parameters of computing which structuresthe problem with the help of evolutionary methods such as chromosomal representation, process of evolution. The process of evolution is the change encoding, selection, crossover and mutation are narrated comparatively.The paper concludes by showing justification of in the inherited traits of a population from one generation to design issues of evolutionary structures. the next. Evolutionary Computation (EC) refers to the computer-based problem solving systems that use Keywords: Evolutionary Algorithm, Evolutionary computational models of evolutionary process. EC uses an Computing, Evolutionary Programming, Evolutionary algorithm to implement evolutionary characteristics in Strategy, Genetic Algorithm, Genetic Programming. solving problems. Such algorithm is known as evolutionary algorithm. Evolutionary characteristics are achieved using I. INTRODUCTION evolutionary computing in which problem is structured Evolutionary Computation (EC) refers to the computer- using chromosome of different types in form of symbols or based problem solving systems that use computational alphanumeric numbers. Evolutionary algorithms are models of evolutionary process.Evolutionary algorithms basically computational models of evolutionary processes. have provided numerous advantages compared to EA are successfullyapplied to varieties of numerous conventional methods. The capability to provide simulation problems from different domains, including intelligent of biological evolution leads evolutionary methods to search, optimization, automatic programming, machine design highly complex mathematical problems. learning, operations research, bioinformatics, and social Evolutionary algorithms provide adaptive nature with systems, engineering systems, health care, medical multimodal characteristics which proves quite favorable in diagnostic, stock and share market, forecasting systems and design of evolutionary systems. EA are successfully applied many more. Fig. 2 represents basic procedure of to varieties of numerous problems from different domains, evolutionary computing [1]. including intelligent search, optimization, automatic programming, machine learning, operations research, Selection bioinformatics, and social systems, engineering systems, Population of Parents medical diagnostic, stock and share market, forecasting Chromosomes systems and many more. Chromosomes The introductory section of the paper explains importance of evolutionary algorithms. The second section of the paper explains the basic procedure of evolutionary Reproduct Replacement algorithm. Differences between traditional search and ion optimization algorithm and evolutionary algorithm are presented in this section. It also describes advantages of evolutionary algorithm. The third section explains Children evolutionary structures such as evolutionary programming, Chromosomes genetic algorithm, genetic programming and Evolutionary Strategies in detail.Evolutionary life cycle and procedural explanation of each evolutionary structure is presented in Fig.1: Basic Elements of Evolutionary Computing this section. The forth section of the paper presents 304 Kunjal Bharatkumar Mankad | IJCSET(www.ijcset.net) | August 2015 | Vol 5, Issue 8,304-308 A. Difference of Traditional optimization Algorithms Vs primary search with an evolutionary algorithm (for Evolutionary Algorithms example fine tuning of weights of a evolutionary There are significant differences observed between GA neural network), or it may involve simultaneous and most of the traditional optimization algorithms as application of other algorithms (e.g., hybridizing summarized by [2,3,4,5,6]: with simulated annealing or tabu search to improve Evolutionary algorithms are ubiquitous in nature the efficiency of basic evolutionary search). while traditional algorithms are static in nature. The evaluation of each solution can be handled in Evolutionary algorithm are specially designed to parallel and only selectionrequires some serial solve mathematical function,which describes the processing. Implicit parallelism is not possible in problem is not known and the values at certain many global optimization algorithms like simulated parametersare obtained from simulations. annealing and Tabu search. Evolutionary algorithm can handle multi-modal Traditional methods of optimization are not robust functions whereas many other to dynamic changes inproblem the environment and optimizationtechniques supports only single model often require a complete restart in order to provide a function; solution (e.g., dynamic programming). In contrast, In traditional optimization, convergence to an evolutionary algorithms can be used to adapt optimal solution depends on the chosen initial solutions to changing circumstance. solution while in EA, due to randomness , initial EA are capable to handle problems which replaces solution is always different; human expertise although human expertise should EA is generic in nature due to objective function be used when it is available, it oftenproves while a traditional algorithm is efficient in solving advantageous for automating problem-solving one problem but the same may not be efficient in routines. solving a different problem; EA works with coding of parameter set rather than III. EVOLUTIONARY STRUCTURES actual value of parameters; A traditional algorithm may not be efficient to handle problems with discrete variables or highly Evolutionary non-linear variables with constraints while EA can Algorithms be robustly applied to problems with any kinds of objective functions, such as nonlinear or step Evolutionary Genetic Evolutionary Genetic functions; because only values of the objective Programming Programming Strategy function for optimization are used to select genes; Algorithm Traditional algorithm can stuck at suboptimal solutions while EA can have less chance to be Fig. 2: Evolutionary Structures trapped by local optima due to characteristics of A. Evolutionary Algorithm crossover and mutation operators. A technique borrowed from the theory of biological evolution that is used to create optimization procedures or B. Advantages of Evolutionary Algorithms methodologies. In nature, evolution is mostly determined EA offers following advantages [6,8]: by natural selection or different individuals competing for A primary advantage of evolutionary computation is resources in the environment. Those individuals that are that it is conceptually simple. The procedure may be better are more likely to survive and propagate their genetic written as difference equation (1) material.EAs fall into the category of “generate and test” x[t + 1] = s(v(x[t])) (1) algorithms.They are stochastic,population-based wherex[t] is the population at time t under a algorithms. EA are based on variation operators representation x, v is a random variation operator, (recombination and mutation) to create the necessary and sis the selection operator . diversity and finally introduce novelty. EA also use Evolutionary algorithm provides representation selection operator whichreduces diversity and provides independent performance, in contrast with other good quality solution.Fig.3 represents pseudo code of an numerical techniques, which might be applicable for evolutionary algorithm [7]. only continuous values or other constrained sets. BEGIN Evolutionary algorithms offer a framework such that INITIALISE population with random candidate solutions; it is comparably easy to incorporate prior knowledge EVALUATE each candidate; about the problem. Incorporating such information REPEAT UNTIL (TERMINATION CONDITION is satisfied) focuses the evolutionary search, yielding a more DO SELECT parent; efficient exploration of the state space of possible RECOMBINE pairs of parents; solutions. MUTATE the resulting offspring; Evolutionary algorithms can also be hybridized with SELECT individuals for the next generation; END other techniques. For ex. This may be as simple as