Discovering justified knowledge from data

Knowledge Discovery from Data (KDD) has become a critical tool for science, commerce and public administration. However, its effectiveness is reduced by the propensity of some key techniques to make many spurious as well as real 'discoveries.' Such false discoveries can be misleading and extremely costly. Standard statistical techniques designed to address this problem do not directly scale-up to the massive numbers of potential discoveries that KDD routinely explores: their power is too low, and their computational requirements too high. This project will develop efficient and effective technologies to control the risk of false discoveries from those susceptible techniques, thereby greatly increasing the value and reliability of KDD.