Bayesian Anomaly Detection

A demonstration of Bayesian anomaly detection for robust parameter inference in the presence of outliers.

Overview

This repository demonstrates how Bayesian anomaly detection can automatically identify and mitigate the effects of outliers in scientific data analysis. The method uses a binary anomaly mask with Bernoulli priors and a piecewise likelihood that gracefully handles both normal and anomalous data points.

Files

📓 Jupyter Notebook

🐍 Python Script

Quick Start

Option 1: Jupyter Notebook

jupyter notebook demo_anomaly_detection.ipynb

Option 2: Python Script

python bayesian_anomaly_detection.py

Key References

• Leeney et al. (2022): Bayesian approach to radio frequency interference mitigation
• Anstey and Leeney (2023): Enhanced Bayesian RFI Mitigation and Transient Flagging Using Likelihood Reweighting

Citation

If you use this code in your research, please cite:

@article{leeney2022bayesian,
  title={Bayesian approach to radio frequency interference mitigation},
  author={Leeney, Samuel and others},
  journal={arXiv preprint arXiv:2211.15448},
  year={2022}
}