Abstract

The quantity of data being collected and stored in the world is growing exponentially, bringing with it the opportunity to generate substantial commercial and societal value from the analysis of that data. As a consequence the field of machine learning, which exploits data to make inferences and predictions, has enjoyed dramatic growth in recent years. At MSR Cambridge we have a substantial programme of research spanning many aspects of machine learning and its applications. In contrast to many traditional approaches to machine learning, we place significant emphasis on the quantification of uncertainty using probabilities. In this talk we will discuss the merits of this approach and illustrate these using a largescale application.