Statistics Seminar - 11/12/26

Abstract:

Many network navigation problems require decisions to be made sequentially while important features of the environment remain uncertain and information is revealed only gradually. We consider this class of problems through extensions of the stochastic obstacle scene framework, in which an agent must navigate toward a destination while accounting for uncertain obstacles, noisy sensing, and costly information acquisition.

I will focus primarily on a spatially correlated version of the problem, where observations at one location provide information about other parts of the network. Spatial dependence is modeled using a Gaussian random field, and Bayesian updating is used to propagate information from noisy, range-limited observations. These evolving beliefs are incorporated into a sequential decision-making framework that combines path-wise decisions, search-space reduction, information-guided learning, and online rollout. I will discuss theoretical and numerical results illustrating how the use of correlation affects traversal performance, robustness, and computational cost.

I will conclude with a complementary extension in which uncertainty arises not only from the environment but also from a strategic opponent. In this setting, a Bayesian traveler learns simultaneously about uncertain network costs and rewards and about the adversary’s behavior, providing a broader perspective on sequential decision-making under uncertainty.

About the Speaker:

Dr. Elvan Ceyhan is the Marguerite Scharnagle Endowed Professor in the Department of Mathematics and Statistics at Auburn University. He received his BS in Mathematics from Koç University, an MS in Statistics from Oklahoma State University, and both an MSE and PhD in Applied Mathematics and Statistics from Johns Hopkins University, followed by a postdoctoral fellowship at JHU’s Center for Imaging Science. Before joining Auburn in 2019, he held faculty positions at Koç University and the University of Pittsburgh and served as Deputy Director of the Statistical and Applied Mathematical Sciences Institute (SAMSI) while a research associate professor at North Carolina State University. He was promoted to full professor at Auburn in 2024. Dr. Ceyhan’s research spans statistical machine learning, graph- and network-based methods, random geometric graphs, and network optimization applied to stochastic obstacle environments and traveler’s problems. His work also includes spatial point pattern analysis and graph-theoretic methods for medical data and image analysis. He has published over 60 peer-reviewed articles, delivered numerous invited presentations, and served in editorial roles for journals such as Computational Statistics and Data Analysis and the Journal of Probability and Statistical Science. An elected member of the International Statistical Institute, he has held leadership roles in the International Association of Statistical Computing and served as president of the ASA Alabama–Mississippi Chapter. His honors include a TWAS Young Affiliate Fellowship, an International Outgoing Fellowship from the European Union, Auburn’s Marie Kraska Award for Excellence in Teaching, and the Phi Kappa Phi Love of Learning Award.

Note:

Contact Prof. JZ at jzhang@math.msstate.edu for additional information.