Black Box Optimization with Data Analysis Export

@phdthesis{citeulike:4742516, abstract = {The main goal of my thesis was to cross the barrier between the fields of optimization and data analysis by applying methods from data analysis to optimization problems. In particular, I applied data analysis techniques to obtain information about black box functions, i.e. functions for which we do not know an algebraic expression. I presented both an algorithm and a reference implementation to solve optimization problems where: * both the objective function and the constraints may be black box functions, * we do not have any gradient or Hessian information for those black box functions, * the functions are assumed to be expensive to compute, thus the number of function evaluations shall be kept as small as possible, using methods from data analysis: * covariance models, * Gaussian mixture models (GMMs) and the Expectation-Maximization (EM) iteration and * ratio-reject. The algorithm is a so-called incomplete global optimizer, i.e. it attempts to find a global solution for the optimization problem, but is unable to guarantee globality. In fact, it cannot even guarantee always finding a local optimum, due to the lack of gradients and any sort of global information. Despite this lack of guarantees, the algorithm performs well in practice.}, author = {Kofler, Kevin}, citeulike-article-id = {4742516}, citeulike-linkout-0 = {http://www.tigen.org/kevin.kofler/bbowda/}, keywords = {continuous, covariance, expectation-maximization, gaussian-mixture, machine-learning, metaheuristic, optimization, stochastic}, posted-at = {2009-06-04 10:05:36}, priority = {3}, school = {Universit\"{a}t Wien}, title = {Black Box Optimization with Data Analysis}, url = {http://www.tigen.org/kevin.kofler/bbowda/}, year = {2007} }

TY - THES ID - citeulike:4742516 L3 - citeulike-article-id:4742516 N2 - The main goal of my thesis was to cross the barrier between the fields of optimization and data analysis by applying methods from data analysis to optimization problems. In particular, I applied data analysis techniques to obtain information about black box functions, i.e. functions for which we do not know an algebraic expression. I presented both an algorithm and a reference implementation to solve optimization problems where: * both the objective function and the constraints may be black box functions, * we do not have any gradient or Hessian information for those black box functions, * the functions are assumed to be expensive to compute, thus the number of function evaluations shall be kept as small as possible, using methods from data analysis: * covariance models, * Gaussian mixture models (GMMs) and the Expectation-Maximization (EM) iteration and * ratio-reject. The algorithm is a so-called incomplete global optimizer, i.e. it attempts to find a global solution for the optimization problem, but is unable to guarantee globality. In fact, it cannot even guarantee always finding a local optimum, due to the lack of gradients and any sort of global information. Despite this lack of guarantees, the algorithm performs well in practice. TI - Black Box Optimization with Data Analysis PB - Universität Wien KW - continuous KW - covariance KW - expectation-maximization KW - gaussian-mixture KW - machine-learning KW - metaheuristic KW - optimization KW - stochastic AU - Kofler, Kevin PY - 2007/// UR - http://www.tigen.org/kevin.kofler/bbowda/ ER -