Online Program Home
My Program

Abstract Details

Activity Number: 260 - SPEED: Topics in Bayesian Analysis
Type: Contributed
Date/Time: Monday, July 30, 2018 : 3:05 PM to 3:50 PM
Sponsor: Section on Statistical Computing
Abstract #332815
Title: Using Modified Competitive Swarm Optimizer to Find D-Optimal Designs for Complicated Logistic Models
Author(s): Zizhao Zhang* and Weng Kee Wong
Companies: UCLA and UCLA
Keywords: Modified Competitive Swarm Optimizer; Locally D-Optimal Design; Pseudo-Bayesian D-Optimal Design; High Dimensional Models; Logistic Models; Optimization

We develop an evolutionary algorithm for solving challenging D-optimal design problems for complex statistical models. The algorithm originates from the Competitive Swarm Optimizer(CSO), which has been shown to be successful in solving many types of large scale optimization problems and easily outperforms other popular evolutionary algorithms. We show that by simply adding some particle mutations at every iteration, the algorithm finds better quality solutions compared to original CSO. We call this version Modified Competitive Swarm Optimizer(MCSO) and demonstrate its superiority using many benchmark test problems. We not only apply this algorithm to find locally D-optimal designs for Logistic models when there are more than five factors and all pairwise interactions but also find pseudo-Bayesian D-optimal designs for Logistic models with three factors and all pairwise interactions under very complicated and various prior combinations. Our work suggests that MCSO is a very promising, powerful and flexible optimization tool, especially for handling complex optimal design problems.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program