Real-time detection of transients in OGLE-IV with application of machine learning

TL;DR

This paper introduces a three-stage hierarchical machine learning system utilizing self-organizing maps for real-time transient detection in OGLE-IV, significantly improving artifact rejection while maintaining high detection efficiency.

Contribution

The paper presents a novel triple-stage hierarchical machine learning approach for artifact filtering in transient detection, enhancing accuracy and efficiency over previous methods.

Findings

Accepts ~97% of real transients

Removes up to ~97.5% of artifacts

Applicable to real-time transient detection

Abstract

The current bottleneck of transient detection in most surveys is the problem of rejecting numerous artifacts from detected candidates. We present a triple-stage hierarchical machine learning system for automated artifact filtering in difference imaging, based on self-organizing maps. The classifier, when tested on the OGLE-IV Transient Detection System, accepts ~ 97 % of real transients while removing up to ~ 97.5 % of artifacts.