Stochastic Simulations Of Search Engines

Computer simulations are widely used in a variety of applications including the military, service industries, manufacturing companies, nuclear power plants, transportation organizations, and manufacturing domain. For example, in nuclear power plants, often computer simulations are used to train personnel on failures and normal operation, study operation plans, support testing of the heart of the nuclear power plant, as well as to evaluate and compare future design plan changes. Other techniques that are used to examine systems, in general, do not have the advantages that computer simulations bring mainly because computer simulations provide cheaper and more realistic results than other approaches. In some cases, computer simulation is the only means to examine a system, like in nuclear power plants, since it is either too dangerous to bring a system under such failure conditions or too costly especially for combat situations to experiment with the system. Also, computer simulations permit us to study systems over a long period of time, to learn from real world past experiences, and to have control over experimental conditions. Actual Internet simulation models are expensive to develop and use in terms of personnel, time and resources. Large memory requirements and slow Response time can prevent companies from considering it as a useful tool. Developing Internet simulations models during training requires models that are very close to reality while their speed is secondary. During testing, speed and reproducibility become the primary issues, which incite us to make the different internal simulation modules as efficient and accurate as possible. Computer simulations have provided companies with the description of the input settings that are needed to produce the optimal best output value for a given set of inputs in a specific domain of study. study. Response surface methodologies using regression models approximations of the computer simulation were the means to achieve computer simulation optimization. As Myers et al. ([104]) stated it in Technometrics,

“There is a need to develop non-parametric techniques in response surface methods (RSM). The use of model-free techniques would avoid the assumption of model accuracy or low-level polynomial approximations and in particular, the imposed symmetry, associated with a second degree polynomial (page 139).”

One possible non-parametric approach is to use artificial neural networks (ANN).