Speaker: Dr. Elvan Ceyhan (Auburn University)

 

Title: Network-Based Learning and Traversal Optimization 

 

 

Abstract: In this talk I will present an overview of my current and recent research program, which centers on (i) network- (or graph-) based statistical machine learning (SML) and (ii) traversal and obstruction optimization in network navigation. Work in (i) aims to address challenges in SML using graph-theoretic tools, e.g., class imbalance, overlap, and overfitting in classification and introduce real-time parameter-free clustering techniques, which, slightly modified, also show promise in outlier detection. The research on (ii) focuses on spatial network optimization for navigating through obstacle-filled environments. The goal is to find the optimal or near-optimal solutions to the two variants of the Stochastic Obstacle Scene (SOS) problem: one involving a navigating agent (NAVA) finding the best path through forbidden regions (Optimal Traversal Path problem), and the other involving an obstacle placing agent inserting obstacles to maximize the NAVA's path (Optimal Obstacle Placement problem). We are working to expand the SOS problem in various directions, develop weight-constrained versions for both variants, and study their solutions. I will also highlight some challenges and future directions in both lines of work.

 

 

Faculty host: Selim  Sukhtaiev