Title: Performant Inference for Infectious Disease Dynamics
Abstract: How big are Big Data and how to wrangle them into Smart Data is a domain-dependent question. Epidemiological studies in infectious disease may scale to hundreds of millions of subjects with billions of health features observed across these subjects. On the other end of the scale, molecular epidemiological studies tracing the evolutionary trajectories of infectious agents and transmission studies involving initial disease outbreaks involve sometimes only tens of observed data samples. However, commonly employed statistical models entertain complex stochastic processes driving these systems. Estimation under the processes usually requires massive missing data augmentation that severely limits their application. In this talk, I outline some recent computational statistics techniques to attack several observed and missing Big Data inference problems. These tools rely on massive parallelization via advancing computing technology to tackle large-scale studies and on tricks from numerical analysis to avoid costly data augmentation for infectious disease transmission, and provide performant inference for these previously intractable problems. Examples draw from the global annual circulation of influenza and the recent Ebola outbreak in West Africa.
LOCATION:TBA STATUS:CONFIRMED DTSTART:20160323T200000Z DTEND:20160323T213000Z END:VEVENT END:VCALENDAR