Finding flares in Kepler and TESS data with recurrent deep neural networks

TL;DR

This paper demonstrates that recurrent neural networks, particularly LSTM-based architectures, can effectively detect stellar flares in space-based photometric data from Kepler and TESS, achieving high accuracy and generalization.

Contribution

The study introduces a neural network approach, especially LSTM RNNs, for automatic flare detection in astronomical data, showing improved accuracy and cross-mission applicability.

Findings

Best RNN architecture used LSTM layers.

Achieved over 80% recall and precision in flare detection.

Successfully generalized from Kepler to TESS data.

Abstract

Stellar flares are an important aspect of magnetic activity -- both for stellar evolution and circumstellar habitability viewpoints - but automatically and accurately finding them is still a challenge to researchers in the Big Data era of astronomy. We present an experiment to detect flares in space-borne photometric data using deep neural networks. Using a set of artificial data and real photometric data we trained a set of neural networks, and found that the best performing architectures were the recurrent neural networks (RNNs) using Long Short-Term Memory (LSTM) layers. The best trained network detected flares over {$5\sigma$ with $\gtrsim80$\% recall and precision} and was also capable to distinguish typical false signals (e.g. maxima of RR Lyr stars) from real flares. Testing the network trained on Kepler data on TESS light curves showed that the neural net is able to generalize and find flares - with similar effectiveness - in completely new data having previously unseen sampling and characteristics.