What's new with the the Department of Mathematics and Statistics?
Colloquium 4/28: Dr. Kalanka Jayalath on Geometrical Pattern Identification Using a Bayesian Paradigm
On Monday April 28th at 3:30, Dr. Kalanka Jayalath will be talking about “Geometrical Pattern Identification Using a Bayesian Paradigm.”
Abstract: Identifying spatially distributed point patterns plays an important role in many scientific areas including pattern recognition, computer vision, image processing and some geological applications. Current methods of identifying conic structures depend solely on algebraic or geometric distances and are known as algebraic or geometric fits respectively. This talk focusses on a novel circle and ellipse fitting technique which elicits a Bayesian philosophy on geometric distance. Statistical methods will be discussed to investigate whether the spatial pattern is reasonably attributable to a circular or elliptical pattern. We compare classical and novel circle fitting methods under various error structures. In particular, we focus on their accuracy of estimates when in the presence of noisy data, a topic that is poorly documented in the literature. Finally, our findings will be applied to a pre-historic archeological site to identify the evident geometrical structure. (Flyer in PDF form)