Real-Time Detection of Anomalies in Large-Scale Transient Surveys (2111.00036v2)

Abstract: New time-domain surveys, such as the Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST), will observe millions of transient alerts each night, making standard approaches of visually identifying new and interesting transients infeasible. We present two novel methods of automatically detecting anomalous transient light curves in real-time. Both methods are based on the simple idea that if the light curves from a known population of transients can be accurately modelled, any deviations from model predictions are likely anomalies. The first modelling approach is a probabilistic neural network built using Temporal Convolutional Networks (TCNs) and the second is an interpretable Bayesian parametric model of a transient. We demonstrate our methods' ability to provide anomaly scores as a function of time on light curves from the Zwicky Transient Facility. We show that the flexibility of neural networks, the attribute that makes them such a powerful tool for many regression tasks, is what makes them less suitable for anomaly detection when compared with our parametric model. The parametric model is able to identify anomalies with respect to common supernova classes with high precision and recall scores, achieving area under the precision-recall curves (AUCPR) above 0.79 for most rare classes such as kilonovae, tidal disruption events, intermediate luminosity transients, and pair-instability supernovae. Our ability to identify anomalies improves over the lifetime of the light curves. Our framework, used in conjunction with transient classifiers, will enable fast and prioritised followup of unusual transients from new large-scale surveys.

Real-Time Detection of Anomalies in Large-Scale Transient Surveys

The paper under review tackles a significant challenge in the domain of time-domain astronomy: the real-time detection of anomalous transient events from the massive datasets expected from upcoming surveys like the Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST). These surveys promise to deliver millions of transient alerts nightly, making traditional manual approaches of inspection infeasible.

Objective and Motivation

The central objective of this research is to develop automated, real-time anomaly detection methods for transient light curves, crucial for prioritizing follow-up observations of astrophysically interesting events. Anomalies in this context are transient light curves that deviate significantly from known populations, offering the potential to discover new astrophysical phenomena.

Methodology

The authors propose two novel methods for anomaly detection based on light curve modeling:

1. Temporal Convolutional Networks (TCNs): This method employs probabilistic neural networks, specifically TCNs, to model expected transient light curves. The detection of anomalies is based on significant deviations from neural network predictions of future fluxes.
2. Bayesian Parametric Modeling: This method utilizes a Bayesian framework using the Bazin function, a parametric model, to fit observed light curves and predict deviations beyond the model expectations.

Notably, both methods focus on analyzing light curves as they evolve over time, thus allowing for a dynamic anomaly detection process.

Numerical Results and Key Findings

The TCN model, with its inherent flexibility in dealing with diverse light curve shapes, surprisingly demonstrated limitations in anomaly detection due to its capability to generalize well across different classes, including anomalous data. Conversely, the Bayesian parametric model was more precise in detecting anomalies, achieving an area under the precision-recall curves (AUCPR) above 0.79 for most rare classes such as kilonovae and tidal disruption events.

Implications and Future Directions

The implications of this research span both practical and theoretical realms:

• Practical Developments: The proposed anomaly detection framework can be integrated into transient alert systems for prioritizing follow-up observations, aiding the discovery of rare and possibly new types of transient phenomena.
• Theoretical Insights: This research prompts further investigation into model flexibility and its trade-offs in the context of anomaly detection, highlighting the necessity for models that balance prediction accuracy with anomaly sensitivity.

Looking forward, future developments could involve enhancing neural network models with constraints to counter their flexibility, applying manifold learning techniques for improved anomaly characterization, and integrating these methods with context-based information from host galaxies for enhanced accuracy.

Conclusion

The paper presents a robust framework for real-time anomaly detection in large-scale astronomical surveys, crucial for addressing the data deluge from new observatories. The methodologies and findings offer a substantial contribution to the field of astrophysical data analysis, providing groundwork for improvements in the detection and paper of transient astronomical events.