Frank, P. D. 1995: Global modeling for optimization. SIAG/OPT Views-and-News. 7, 9--12.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....and become trapped in local minimizers created by high frequency oscillations. In order to develop effective algorithms for such applications, we restrict attention to derivative free methods for numerical optimization. Second, complicated computer simulations are often expensive to perform. Frank (1995) suggested that one must address problems in which a typical function evaluation costs several hours of supercomputer time. For example, Booker (1996) studied an aeroelastic and dynamic response simulation of a helicopter rotor blade for which a single function evaluation requires approximately ....

....and potentially better ways of implementing this traditional practice. The prescription that seems to be gaining some currency in the engineering community was proposed by Welch and Sacks (1991) following current convention, we refer to it as DACE (Design and Analysis of Computer Experiments) Frank (1995) offered an optimizer s perspective on this methodology, suggested that the minimalist approach of minimizing a single f is not likely to yield satisfactory results, and proposed several sequential modeling strategies as alternatives. Booker (1996) studied several industrial applications of ....

[Article contains additional citation context not shown here]

Frank, P. D. (1995). Global modeling for optimization. SIAG/OPT Views-and-News, (7):9--12.

....and geostatistics as described by Cressie in [12, pages 1 26] and [13] The term kriging was first used in the work of Matherton [14] who attributed the original development of the statistical techniques to mining engineer D. G. Krige. Recent studies by researchers at Boeing including Frank [15] and Booker [16] have employed DACE modeling methods in engineering optimization problems. The DACE modeling methods differ from the polynomial modeling methods in that the DACE models can capture oscillatory (multimodal) response trends whereas quadratic polynomials are by definition unimodal. ....

Frank, P. "Global Modeling for Optimization," SIAG/OPT Views-and-News, 7, 9--12 (1995).

....it approximates the objective function over the entire feasible region, recalibration of the approximation is made using one new objective value at a time, as these values are produced by the optimization procedure. In contrast to MAGS, a zoom in method for local refinement was proposed in [18]. This strategy uses the existing approximation to determine an interesting subregion of the design space for further exploration. The optimization process is halted, additional function values are obtained in the subregion, and a new approximation is formed. The expectation is that the ....

Frank, P. D. 1995: Global modeling for optimization. SIAG/OPT Views-and-News. 7, 9--12.

....it approximates the objective function over the entire feasible region, recalibration of the approximation is made using one new objective value at a time, as these values are produced by the optimization procedure. In contrast to MAGS, a zoom in method for local refinement was proposed in [18]. This strategy uses the existing approximation to determine an interesting subregion of the design space for further exploration. The optimization process is halted, additional function values are obtained in the subregion, and a new approximation is formed. The expectation is that the ....

Frank, P. D. 1995: Global modeling for optimization. SIAG/OPT Views-and-News. 7, 9--12.

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