CLOE: Christoffel Loss Autoencoder for Anomaly Detection

Short summary

CLOE combines an autoencoder for dimensionality reduction with a Christoffel Function-based anomaly detector applied in latent space, addressing the poor scalability of Christoffel methods to high-dimensional data. A novel loss function leverages the Christoffel Function to guide the autoencoder toward representations that better capture the normal data distribution's support. Experiments on high-dimensional tabular benchmarks show CLOE outperforms existing methods while preserving the lightweight, low-tuning advantages of Christoffel Function approaches.

• •CLOE pairs an autoencoder with Christoffel Function-based detection in latent space for high-dimensional anomaly detection
• •Novel Christoffel-guided loss aligns representation learning with anomaly detection objectives
• •Outperforms existing methods on tabular benchmarks while requiring minimal hyperparameter tuning

Generated with AI, which can make mistakes.