Surprisal-Based Anomaly Detection in Sparse High-Dimensional Data

NetSciX 2026 Conference Submission34 Authors

Keywords: surprisability, missing commonalities, high-dimensional data, anomaly detection

TL;DR: Learning via Surprisability (LvS) detects anomalies in high-dimensional data by identifying both overrepresented features and "missing commonalities" (expected features that are absent), achieving superior performance compared to traditional methods.

Submission Number: 34