Abstract

Abstract: The Covid-19 pandemic is testing the ability to use widespread personal sensor data for health-related applications, particularly in the case of contact tracing. This talk will present some of the first work that used large-scale mobile phone data as a proxy for physical proximity in simulating epidemics. Two works will be presented on digital epidemiology, the first on digital contact tracing and the second on Bluetooth networks and the ability to predict whom will become infected next over time. Finally, another line of research related to Covid-19 on x-ray data will be presented that uses cascade transfer learning from ImageNet and has been validated on various applications including sensor-driven Human Activity Recognition.

Bio: Kate Farrahi is an assistant professor in computing at the University of Southampton. She works in the vision, learning and control group. The main focus of her research is machine learning, particularly deep learning, often applied to health applications. In 2014, together with her collaborators she led some work on digital contact tracing, which appears to be the first paper on the subject. In addition to epidemics, other applications of her research include mood predictions from sensor data, and Parkinson’s disease prediction from typing data. Prior to joining Southampton, she was a research assistant at Idiap Research Institute and obtained her PhD in Computer Science from the Swiss Federal Institute of Technology in Lausanne (EPFL).