Oral Defense of Doctoral Dissertation – Computational Sciences and Informatics – New Methods Of Spectral-Density Based Graph Construction And Their Application To Hyperspectral Image Analysis – Jeffrey Stevens
Notice and Invitation
Oral Defense of Doctoral Dissertation
Doctor of Philosophy in Computational Sciences and Informatics
Department of Computational and Data Sciences
College of Science
George Mason University
Bachelor of Science, State University of New York at Buffalo, 1992
Master of Science, State University of New York at Buffalo, 1994
New Methods Of Spectral-Density Based Graph Construction And Their Application To Hyperspectral Image Analysis
Monday, November 20, 2017, 2:00 -4:00 p.m.
Research Hall, Room 162
All are invited to attend.
Dr. John J. Qu, Dissertation Director
Dr. Ronald G. Resmini
Dr. Robert S. Weigel
Dr. Juan R. Cebral
The past decade has seen the emergence of many hyperspectral image (HSI) analysis algorithms based on graph theory and derived manifold-coordinates. Yet, despite the growing number of algorithms, there has been limited study of the graphs constructed from spectral data themselves. Which graphs are appropriate for various HSI analyses—and why? This research aims to begin addressing these questions as the performance of graph-based techniques is inextricably tied to the graphical model constructed from the spectral data.
We advance the performance of all graph- and manifold-based analyses by providing guidelines developed from rigorous study to ensure the starting graphs are of higher quality, and demonstrate the effectiveness of shared nearest neighbors (SNN) and mutual proximity (MP) as alternative methods of graph construction not limited by assumptions of current techniques. In developing construction guidance, we consider the impacts of generating graphs in higher dimensional HSI data and present the first study of its kind related to concentration of measure and hubness in graphs of HSI data indicating the types of analyses that can be impacted by these phenomena. To conclude, we introduce new algorithms based on SNN and MP for classification and anomaly detection using a Schroedinger Eigenmaps formulation and compare performance to leading techniques.
A copy of Jeff’s dissertation is available for examination from Karen Underwood, Computational and Data Sciences, 373 Research Hall. The dissertation is available to read only within the Department and cannot be taken out of the department or copied.