Anomaly Detection: Algorithms, Explanations, Applications
Anomaly detection is important for data cleaning, cybersecurity, and robust AI systems. This talk will review recent work in our group on (a) benchmarking existing algorithms, (b) developing a theoretical understanding of their behavior, (c) explaining anomaly “alarms” to a data analyst, and (d) interactively re-ranking candidate anomalies in response to analyst feedback. Then the talk will describe two applications: (a) detecting and diagnosing sensor failures in weather networks and (b) open category detection in supervised learning.
Anomaly Detection Algorithms and Techniques for Real-World Detection Systems
Finding outliers in a dataset is a challenging problem in which traditional analytical methods often perform poorly. As a result, researchers have developed special algorithms for detecting anomalies. In this talk, I will take about three different families of anomaly detection algorithms: Density-based methods, data streaming methods, and time series methods. I will cover both the mathematical and statistical theory behind these algorithms and provide code implementations. Afterwards, I will discuss useful tips I have learned while implementing threat detection systems in practice.