Different Methods of Software Optimization - Essay Example

Proposal In the current global environment, digital technology and the increase in volume of online transmissions requires the availability of considerable amounts of bandwidth, further exacerbated by the lack of availability of storage space on computers. Software optimization offers an excellent solution to both problems, providing for more efficient usage with fewer resources. This research study will focus upon software optimization through a comparison of two different methods of software optimization: (a) genetic algorithms and (b) distribution algorithms. Research goals: There are various methods which have been used for software optimization. Genetic algorithms are a method for solving problems, which use the same techniques of selection and mutation that occurs in mother nature, in order to arrive at a solution to a problem. In order to derive a genetic algorithm, potential solutions to a problem need to be developed by encoding; for instance by using binary bit strings. Another method used is the estimation of distribution algorithms for software optimisation, whereby a probability model is maintained for each generation of a problem situation. While there are studies applying each of these software optimization methods separately, the number of studies where both methods are compared in the same assessment are not as plentiful. This study seeks to carry out a comparison of both these software optimization methods in one particular study. Both these methods will be compared and an estimation of the efficacy of each method in terms of software optimization capacity will be made on the basis of the comparison. This study will rely primarily on two major research studies which are detailed below, which form a close parallel to the subject of this research study. Literature review: The most relevant studies that would apply in the context of this research proposal are those showing different methods of software optimization. Through the use of parallel and distributing processing, multi thread techniques have been shown to provide better solutions than sequential options (Cruz and Pelta, 2009). Under the multi thread option, each solver thread represents a particular optimization algorithm and a coordinator collects performance information on the solvers and then sends them instructions on how behaviour is to be altered. On the algorithmic approach, metaheuristics can be successfully applied to complex and difficult combinations of optimization programs. In the study carried out by Cruz and Pelta (2009), the most basic ingredients of soft computing were used, i.e, through fuzzy sets and fuzzy rules, depending upon the location, i.e, by focusing upon the p-median, where the combinatorial optimization problem occurred. The first heuristic was the application of a Fuzzy Adaptive Neighbourhood Search, where various neighbourhoods are available and different acceptable solutions are generated. Both sequential and cooperative methods were used in terms of computational effort. The study by Cruz and Pelta (2009) shows that in obtaining a successful solution to the p-median problem, the use of FANS and a cooperative strategy were found to be beneficial. Discrete Optimization: In the context of discrete optimization, it may best be described as a collection of problem situations, each of which is specified by a pair of variables, i.e, s which is a set of finite candidate configurations and f is the cost or the objective function that is assessed using a formula (Hao, no date). Many theoretical and practical problems can be solved; in the context of optimization, these could include problems of general integer programming, bandwidth minimization or linear arrangement. It may also cover other problems such as environment problems, renewable energy, communications and productivity in manufacturing and service sectors, to provide just a few examples. Therefore, as a part of the optimization process, there are two aspects – exploitation and exploration. The former refers to the ability of a method to intensively examine specific areas , while exploration refers to the method of diversifying a search in order to discover new search areas. If discrete optimization is to be successful, then it must be able to apply both areas, i.e, exploitation as well as exploration and conciliate them both. In achieving this objective, Hao (no date) has proposed the use of memetic algorithms, i.e, hybrid search methods that are based upon both population based search frameworks as well as neighbourhood based local search frameworks. The former would involve the use of genetic and other evolutionary algorithms, while the latter would include Tabu search and simulated annealing. The MA has some basic components such as a population of individuals, a crossover and new offspring created by blending two or more existing solutions. The initial population moves through repeat cycles of evolution, each of which consists of four sequential steps. The first step is the selection of parents, the next is the combination of the parents to generate offspring, the third is the local improvement of the offspring and the final stage is the updating of the population. In terms of performance, an MA performance depends upon the design of its two search components, i.e, combination and local improvement operators, with the design also incorporating problem-specific knowledge of the given problem to ensure effective levels of exploitation and exploration. Method: This study proposes to apply both the methods, i.e, algorithmic method of software optimization and the method of discrete optimization, in order to carry out a comparison of the two and arrive at a conclusion as to which one is the better one. At the outset, the p-median problem will be examined, by determining a set L of x no: of potential locations for software centres and another set U containing the locations of y number of users. This will be used to devise an x X y or W matrix, which will contain the costs incurred to satisfy the demand of user a from location b. The goal would be to select from c number of columns from the set W, such that the sum of the minimum coefficients in each line that falls within these columns being the smallest. Once the p-median has been determined, this will be used as the basis to develop FANS. In determining discrete optimization, a memetic algorithm template will be developed, into which the input will be the size of the population in the locations where search optimization is to examined, with the output being the best solution found for optimization. The fitness of each individual will be evaluated and the best solution in terms of the fitness of each solution will be developed. On the basis of this, the best solution will be determined and then compared with the results obtained from the application of fuzzy valuation. This comparative assessment would provide an idea of which search optimization technique would be the best. References: *Cruz, Carlos and Pelta, David, 2009. “Soft computing and cooperative strategies for optimization”, Applied Soft Computing, 9:30-28 *Hao, Jin-Kao, No date. “Memetic algorithms for discrete optimization”, Read More