PACWrap: Semi-Supervised PAC Anomaly Detector
Faculty: Insup Lee, Edgar Dobriban, and Oleg Sokolsky
Opportunity: Design a framework that bounds the error rates in anomaly detection.
Challenge: How to ensure Probably Approximately Correct (PAC) guarantee on both the false-negative (FNR) and false-positive (FPR) detection rates for any anomaly detector.
- Without the need for labeled anomaly data during training