Spotting anomalies with Privileged Information

SPI is a new ensemble approach that leverages privileged information—data available only for training examples—for unsupervised anomaly detection.

SPI constructs a number of frames/fragments of knowledge (i.e., density estimates) in the privileged space and transfers them to the anomaly scoring space through "imitation" functions that use only the partial information available for test examples.

SPI enables resource-frugal, early, and preventive detection of anomalies as demonstrated on benchmark and real-world datasets.

Shubhranshu Shekhar and Leman Akoglu. Incorporating Privileged Information to Unsupervised Anomaly Detection. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases (ECML-PKDD), 2018.