Zimmerman, D. C., Genetic Algorithms for Navigating Expensive and Complex Design Spaces, Final Report for Sandia National Laboratories contract AO-7736 CA 02, Sept. 1996.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....[34] The algorithm used in the kriging process generates a C continuous surface that exactly interpolates the data values. Artificial Neural Networks: An implementation of the stochastic layered perceptron neural network developed by Prof. D. C. Zimmerman of the University of Houston [66]. This neural network method is intended to have a lower training (fitting) cost than typical neural networks. Multivariate Adaptive Regression Splines (MARS) Software developed by Prof. J. H. Friedman of Stanford University [25] The MARS method creates a C continuous patchwork of splines in ....

....optimization. x ( b r 1 f be ( b Qq 1 . q n , f x ( 14.6 Artificial Neural Network (ANN) Models The ANN surface fitting method in DAKOTA employs a stochastic layered perceptron (SLP) artificial neural network based on the direct training approach of Zimmerman [66]. The SLP ANN method is designed to have a lower training cost than traditional ANNs. This is a useful feature for SBO and OUU where new ANNs are constructed many times during the optimization process (i.e. one ANN for each response function, and new ANNs for each optimization iteration) The ....

Zimmerman, D. C., Genetic Algorithms for Navigating Expensive and Complex Design Spaces, Final Report for Sandia National Laboratories contract AO-7736 CA 02, Sept. 1996.

....in this study employed four surrogate modeling techniques. These were: 1) kriging spatial interpolation [20, 21] 2) quadratic polynomial regression (QuadPoly) 22] 3) multivariate adaptive regression splines (MARS) 23] and (4) stochastic layered perceptron artificial neural networks (ANN) [24]. The kriging, MARS, and ANN methods do not assume a particular trend in the data. That is, these three surrogate modeling methods can capture arbitrary variations in a given data set. In contrast, the quadratic polynomial regression assumes that the data trends can be modeled using second order ....

Zimmerman, D., "Genetic Algorithms for Navigating Expensive and Complex Design Spaces," Final Research Report prepared for Sandia National Laboratories (technical contact M. Eldred), 1996. 11 American Institute of Aeronautics and Astronautics

....Interface 98 This capability is new and evolving. Additional details will be provided in future documentation releases. The ANN Approximation Interface The ANN Approximation Interface uses a layered perceptron artificial neural network based on the direct training approach of Zimmerman ([Zimmerman, D.C. 1996]) Following evaluation of the DDACE sample points with the actual model, the ANN weights are computed with an SVD decomposition. This capability is new and evolving. Additional details will be provided in future documentation releases. User s Instructions Exploiting Parallelism Parallelism ....

Zimmerman, D.C., 1996 Genetic Algorithms for Navigating Expensive and Complex Design Spaces, Final Report for Sandia National Laboratories contract AO-7736 CA 02, Sept. 1996.

....to most common UNIXbased workstations including Sun, SGI, DEC, IBM, and LINUX based PCs. The DAKOTA software provides much of the infrastructure needed to implement an SAO strategy. DAKOTA has a library of surrogate modeling methods including polynomial regression [10] artificial neural networks [11], multivariate regression splines [12] and kriging interpolation [13] Data sampling methods in DAKOTA include Latin hypercube sampling [14] orthogonal array sampling [15] and pure random sampling. 3 Trust Region Methods 3.1 Background Consider a nonlinear inequality constrained problem ....

Zimmerman, D., "Genetic Algorithms for Navigating Expensive and Complex Design Spaces," Final Research Report prepared for Sandia National Laboratories, (technical contact: M. Eldred) , 1996.

No context found.

Zimmerman, D.C. "Genetic Algorithms for Navigating Expensive and Complex Design Spaces," Final Research Report for Sandia National Laboratories contract A0-7736 CA02, (technical contact: M.S. Eldred), 1996.

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