Unsupervised Dictionary Learning for Anomaly Detection
This work addresses anomaly detection for domains requiring unsupervised and online methods, but it appears incremental as it builds on existing dictionary learning techniques.
The paper tackles anomaly detection in applications like anti-money laundering by proposing an unsupervised dictionary learning approach, achieving low false positive rates and demonstrating results with their semi-supervised online algorithm TODDLeR.
We investigate the possibilities of employing dictionary learning to address the requirements of most anomaly detection applications, such as absence of supervision, online formulations, low false positive rates. We present new results of our recent semi-supervised online algorithm, TODDLeR, on a anti-money laundering application. We also introduce a novel unsupervised method of using the performance of the learning algorithm as indication of the nature of the samples.