Sonia Petrone Acknowledged with an IMS Medallion Lecture

Sonia Petrone (Department of Decision Science) has been honored by the Institute of Mathematical Statistics (IMS) which has invited her to deliver a Medallion Lecture at the 2018 IMS Annual Meeting on Probability and Statistics (Vilnius, Lithuania, 2-6 July 2018).

The speakers of the Medallion Lectures (so named because the Lecturers receive a medallion in a brief ceremony preceding their speech) are selected by the IMS Special Lectures Committee as an acknowledgement to their contribution to the field.

Prof. Petrone will deliver a lecture on predictive learning. Traditionally, in the inferential approach to statistics, the focus is on the model's parameters, in the predictive approach it is on future events. "In the last years", Prof. Petrone explains, "many predictive Bayesian methods have been developed and applied to different fields, such as text analysis, genetics, population dynamics, ecology, economics, finance and so on. It's a continuing challenge", Prof. Petrone goes on, "with more complex data – images and videos, for instance – on many different dimensions – time, space, graphs, networks – and with many more interdependencies".

Scholars and practitioners of predictive learning, thus, face not only statistical but also computational complexity and, in order to overcome this latest issue, have developed a series of approximation methods that allow to manage the models. "These methods are good and seem to work well", Prof Petrone concludes, "but they often lack clear theoretical foundations and it's hard to measure how good they really are. I'm trying to developed theory-based methods that couple good prediction with the possibility to measure the quality of the approximation".

Established in 1935, IMS has 4,000 members all over the world and publishes some of the most prestigious peer-reviewed journals of the field, including Annals of Statistics, Annals of Probability and Statistical Science.