Semi-parametric Robust Event Detection for Massive Time-Domain Databases

TL;DR

This paper introduces a semi-parametric, robust, and scalable method for detecting isolated events in massive, irregularly-sampled time-series data, specifically tailored for astronomical surveys, improving efficiency and accuracy.

Contribution

The authors develop a novel semi-parametric Bayesian approach that effectively handles non-linear trends and non-Gaussian noise in large-scale astronomical time-series data.

Findings

Achieved over 100-fold reduction in candidate events

Successfully applied to 87.2 million sources from EROS-2

Produced high-quality features for further analysis

Abstract

The detection and analysis of events within massive collections of time-series has become an extremely important task for time-domain astronomy. In particular, many scientific investigations (e.g. the analysis of microlensing and other transients) begin with the detection of isolated events in irregularly-sampled series with both non-linear trends and non-Gaussian noise. We outline a semi-parametric, robust, parallel method for identifying variability and isolated events at multiple scales in the presence of the above complications. This approach harnesses the power of Bayesian modeling while maintaining much of the speed and scalability of more ad-hoc machine learning approaches. We also contrast this work with event detection methods from other fields, highlighting the unique challenges posed by astronomical surveys. Finally, we present results from the application of this method to 87.2 million EROS-2 sources, where we have obtained a greater than 100-fold reduction in candidates for certain types of phenomena while creating high-quality features for subsequent analyses.