Accounting for Changes in the Distribution of Data

Accomodating changes to the distribution of data in classification and inference procedures is an important

problem for statistics. For instance, statistical predictions often rely on the assumption that future data will follow the same distribution as past data, and predictions may fail when there is a change to the data generating process. In other cases, detecting changes to the distribution of data is of scientific interest itself. For example, when monitoring a patient's brain activity, a change may have serious health implications; and in cellular biology, characterizing differences between biological conditions is a key goal for experiments.