Y. Jin, M. Olhofer, and B. Sendhoff, "Managing approximate models in evolutionary aerodynamic design optimization," in Proceedings of IEEE Congress on Evolutionary Computation, Vol. 1, Seoul, Korea, 2001, pp. 592--599.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....of evaluated solutions and their prediction quality improves with the growing number of evaluated solutions in the optimization process. Models can be incorporated into an evolutionary optimization in various ways. As a first approach, new solutions (o#spring) can be (pre )evaluated by the model [3]. The pre evaluation can be used to indicate promising solutions. It is far from clear however which of the new solutions should be evaluated on the objective function and the fraction of evaluated solutions varies in literature between 10 [4] and 50 [5] In addition, the inexact pre evaluation ....

Jin, Y., Olhofer, M., Sendho#, B.: Managing approximate models in evolutionary aerodynamic design optimization. In: Proceedings of IEEE Congress on Evolutionary Computation, Vol. 1, Seoul, Korea. (2001) 592--599

....computation the evaluation of the individual s fitness is the most time consuming component of the optimization. One attempt to reduce this time is to substitute the original fitness function at least in some generations by an approximate model with a much lower computational cost [7] In [6], a framework for evolutionary optimization using approximate models with application to design optimization has been proposed. In this framework, the approximate model is combined with the original fitness function to control the evolutionary process, i.e. to decide to which proportion the ....

....the individuals are evaluated by means of the original fitness function, in the remaining ones by means of the approximate model. During the first # generations, the model output is compared with the original fitness function to adapt the value of #; this procedure is denoted as evolution control [6]. Of course, the accuracy of the approximate model strongly determines the e#ciency of this approach. As it is quite unlikely to find an accurate model for the whole fitness landscape, it seems to be more promising only to model the local vicinity of the actual population. Therefore, the genomes ....

Yaochu Jin, Markus Olhofer, and Bernhard Sendho #. Managing approximate models in evolutionary aerodynamic design optimization. In Proceedings of the 2001.

....Both measures use the information which is already available within the population to estimate the variance. Thus, no additional fitness evaluations are necessary, which is very important when fitness evaluations is computationally expensive, such as in aerodynamic design optimization problems [5]. In Section two, we will briefly review some of the expectation based approaches to searching for robust optimal solutions. The multiobjective optimization algorithm used in this paper, the dynamic weighted aggregation method proposed in [6, 7] is described in Section 4. Simulation results on ....

Y. Jin, M. Olhofer, and B. Sendho#. Managing approximate models in evolutionary aerodynamic design optimization. In IEEE Congress on Evolutionary Computation, volume 1, pages 592--599, Seoul, South Korea, 2001.

....research topics are discussed in Section 7. 2 Motivations So far, approximation of the tness function in evolutionary computation has been applied mainly in the following cases. The computation of the tness is extremely timeconsuming. One good example is structural design optimization [30, 43, 44, 51, 77, 59, 37, 55]. In aerodynamic design optimization, it is often necessary to carry out computational uid dynamics (CFD) simulations to evaluate the performance of a given structure. A CFD simulation is usually computationally expensive, especially if the simulation is 3 dimensional, which takes over ten hours ....

....than do them randomly. However, the reduction of tness evaluations may not be signi cant. Use of approximate tness models through tness evaluations. In most research, the approximate model has been directly used in tness evaluations in order to reduce the number of tness calculation [44, 56, 57, 48, 64, 65, 20, 21, 51, 12, 59, 23, 36, 37, 38, 27, 22]. Di erent approximate models, including polynomials, kriging models and neural networks have been applied. An interesting idea in [22] is that a con dence interval for the tness estimation is calculated to modify the model prediction so that the search in unexplored regions is encouraged. It ....

[Article contains additional citation context not shown here]

Y. Jin, M. Olhofer, and B. Sendho. Managing approximate models in evolutionary aerodynamic design optimization. In Proceedings of IEEE Congress on Evolutionary Computation, volume 1, pages 592{ 599, May 2001.

....computation the evaluation of the individual s fitness is the most time consuming component of the optimization. One attempt to reduce this time is to substitute the original fitness function at least in some generations by an approximate model with a much lower computational cost [7] In [6], a framework for evolutionary optimization using approximate models with application to design optimization has been proposed. In this framework, the approximate model is combined with the original fitness function to control the evolutionary process, i.e. to decide to which proportion the ....

....the individuals are evaluated by means of the original fitness function, in the remaining ones by means of the approximate model. During the first # generations, the model output is compared with the original fitness function to adapt the value of #; this procedure is denoted as evolution control [6]. Of course, the accuracy of the approximate model strongly determines the e#ciency of this approach. As it is quite unlikely to find an accurate model for the whole fitness landscape, it seems to be more promising only to model the local vicinity of the actual population. Therefore, the genomes ....

Yaochu Jin, Markus Olhofer, and Bernhard Sendho #. Managing approximate models in evolutionary aerodynamic design optimization. In Proceedings of the 2001.

No context found.

Y. Jin, M. Olhofer, and B. Sendhoff, "Managing approximate models in evolutionary aerodynamic design optimization," in Proceedings of IEEE Congress on Evolutionary Computation, Vol. 1, Seoul, Korea, 2001, pp. 592--599.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC