Evolutionary algorithms (EAs) have received increasing interests both in the academy and industry. One main difficulty in applying EAs to real-world applications is that EAs usually need a large number of fitness evaluations before a satisfying result can be obtained. However, fitness evaluations are not always straightforward in many real-world applications. Either an explicit fitness function does not exist, or the evaluation of the fitness is computationally very expensive. In both cases, it is necessary to estimate the fitness function by constructing an approximate model. In this paper, a comprehensive survey of the research on fitness approximation in evolutionary computation is presented. Main issues like approximation levels, approximate model management schemes, model construction techniques are reviewed. To conclude, open questions and interesting issues in the field are discussed.

In this paper we propose two novel approaches for solving constrained multi-objective optimization problems using steady state GAs.

Abstract—Optimization is often a critical issue for most system design problems. Evolutionary Algorithms are population-based, stochastic search techniques, widely used as efficient global optimizers. However, finding optimal solution to complex high dimensional, multimodal problems often require highly computationally expensive function evaluations and hence are practically prohibitive. The Dynamic Approximate Fitness based Hybrid EA (DAFHEA) model presented in our earlier work [14] reduced computation time by controlled use of meta-models to partially replace the actual function evaluation by approximate function evaluation. However, the underlying assumption in DAFHEA is that the training samples for the meta-model are generated from a single uniform model. Situations like model formation involving variable input dimensions and noisy data certainly can not be covered by this assumption. In this paper we present an enhanced version of DAFHEA that incorporates a multiple-model based learning approach for the SVM approximator. DAFHEA-II (the enhanced version of the DAFHEA framework) also overcomes the high computational expense involved with additional clustering requirements of the original DAFHEA framework. The proposed framework has been tested on several benchmark functions and the empirical results illustrate the advantages of the proposed technique. Keywords—Evolutionary algorithm, Fitness function, Optimization, Meta-model, Stochastic method.

Abstract- The focus of this work is to investigate the effects of applying the singular value decomposition (SVD), a linear algebra technique, to the domain of Genetic Algorithms. Empirical evidence, concerning document comparison, suggests that the SVD can be used to model information in such a way that provides both a saving in storage space and an improvement in information retrieval. It will be shown that these beneficial properties can be extended to many other different types of comparison as well. Briefly, vectors representing the genes of individuals are projected into a new low-dimensional space, obtained by the singular value decomposition of a gene-individual matrix. The information about what it means to be a good or bad individual serves as a basis for qualifying candidate individuals for reinsertion into the next generation. Positive results from different approaches of this application are presented and evaluated. In addition, several possible alternative techniques are proposed and considered. 1

This item was submitted to Loughborough’s Institutional Repository by the

In this paper we propose a novel approach for solving constrained multi-objective optimization problems using a steady state GA and reduced models. Our method called Objective Exchange Genetic Algorithm for Design optimization (OEGADO) is intended for solving real-world application problems that have many consmints and very small feasible regions. OEGADO runs several GAs concurrently with each GA optimizing one objective and exchanging information about its objective with others. Empirical results in benchmark and engineering design domains are presented. A comparison between OEGADO and Non-Dominated Sorting Genetic Algorithm-II (NSGA-II) shows that OEGADO performed better than NSGA-II for difficult problems, and found Pareto-optimal solutions in fewer objective evaluations. The results suggest that our method may be better for solving real-world application problems wherein the objective computation time is large.

In this paper we compare several machine learning techniques for evaluating external skeletal fixation proposals. We experimented in the context of dog bone fractures but the potential applications are numerous. Decision trees tend to give both binary and multiple class prediction quickly and accurately. The classifier system method does worse due to the small size of the data set and missing values. The use of artificial Neural Networks is promising, although it takes more time in training. A Genetic Algorithm is also employed to find the best structure of the Neural Network. Experimental results for the different methods are presented and compared. 1.

A reduced model technique based on polynomial approximations is developed in order to increase the rate of convergence of an evolution strategy (ES) when solving a non-destructive evaluation problem. The inverse problem investigated consists of identifying the geometry of discontinuities in a conductive material from Cauchy data measurements taken on the boundary. In this study we use polynomial approximation models in order to increase the rate of convergence of the optimisation algorithm and to e#ciently detect, from a computational time point of view, a subsurface cavity, such as a circle. The algorithm developed by combining evolution strategies and polynomial approximations is found to be a robust, fast and e#cient method for detecting the size and location of subsurface cavities.

Evolutionary Algorithms are population-based, stochastic search techniques, widely used as efficient global optimizers. However, many real life optimization problems often require finding optimal solution to complex high dimensional, multimodal problems involving computationally very expensive fitness function evaluations. Use of evolutionary algorithms in such problem domains is thus practically prohibitive. An attractive alternative is to build meta models or use an approximation of the actual fitness functions to be evaluated. These meta models are order of magnitude cheaper to evaluate compared to the actual function evaluation. Many regression and interpolation tools are available to build such meta models. This paper briefly discusses the architectures and use of such meta-modeling tools in an evolutionary optimization context. We further present two evolutionary algorithm frameworks which involve use of meta models for fitness function evaluation. The first framework, namely the Dynamic Approximate Fitness based Hybrid EA (DAFHEA) model [14] reduces computation time by controlled use of meta-models (in this case approximate model generated by Support Vector Machine regression) to partially replace the actual function evaluation by approximate function evaluation. However, the underlying assumption in DAFHEA is that the training samples for the metamodel are generated from a single uniform model. This does not take into account uncertain scenarios involving noisy fitness functions. The second model, DAFHEA-II, an enhanced version of the original DAFHEA framework, incorporates a multiple-model based learning approach for the support vector machine approximator to handle noisy functions [15]. Empirical results obtained by evaluating the frameworks using several benchmark functions demonstrate their efficiency

stochastic search techniques, widely used as efficient global optimizers. However, many real life optimization problems often require finding optimal solution to complex high dimensional, multimodal problems involving computationally very expensive fitness function evaluations. Use of evolutionary algorithms in such problem domains is thus practically prohibitive. An attractive alternative is to build meta models or use an approximation of the actual fitness functions to be evaluated. These meta models are order of magnitude cheaper to evaluate compared to the actual function evaluation. Many regression and interpolation tools are available to build such meta models. This paper briefly discusses the architectures and use of such meta-modeling tools in an evolutionary optimization context. We further present two evolutionary algorithm frameworks which involve use of meta models for fitness function evaluation. The first framework, namely the Dynamic Approximate Fitness based Hybrid EA (DAFHEA) model [14] reduces computation time by controlled use of meta-models (in this case approximate model generated by Support Vector Machine regression) to partially replace the actual function evaluation by approximate function evaluation. However, the underlying assumption in DAFHEA is that the training samples for the metamodel are generated from a single uniform model. This does not take into account uncertain scenarios involving noisy fitness functions. The second model, DAFHEA-II, an enhanced version of the original DAFHEA framework, incorporates a multiple-model based learning approach for the support vector machine approximator to handle noisy functions [15]. Empirical results obtained by evaluating the frameworks using several benchmark functions demonstrate their efficiency