Designs for Cross-Validation and Stochastic Optimization

Cross-validation and stochastic optimization (i.e., optimization involving integrals) are critical for solving a large array of estimation problems in statistics, applied mathematics and computer sciences. This talk is devoted to showcasing the importance of design, i.e., efficient data collection, for improving these two procedures. Specific topics include using space-filling designs to enhance sample average approximations for various stochastic optimization problems and employing sliced space-filling designs to significantly reduce the variability of cross-validation in estimation of the error rate of a classification rule.

Event Topic:

Computational Mathematics & Statistics