Poster Abstract

P10.13 Meng Zhu (college of information science and technology, Beijing normal university)

Theme: Data processing pipelines

Detection of GWAC Abnormal Light Transform Based on Sparse Autoencoder

The Geographic Wide Angle Camera Array (GWAC) is an important ground-based
observation device for the Sino-French astronomical satellite project. It can get millions of the
light curves with 15 seconds of time resolution per night, which is beneficial to do research
in observation of gravitational microlensing (an important probe observation of exoplanets),
flare star, unknown transient objects (Gravitational-Wave Bursts). These transient sources
rarely show up and always have anomalous properties on light and appear very few times.
It is a challenging job to detect the celestial bodies with abnormal light from the observed
light curves. Thus, to improve the adverse conditions of light curves manual interpretation
such as low efficiency and inevitable errors or omissions, this paper presents a method based
on Sparse Autoencoder (SAE) for light curves abnormal detection. In order to demonstrate
the effectiveness of our proposed algorithm, we do some experiments and compare with K-
means in false positive rate and the elapsed time. We identified 3 unknown variability types
and a few individual outliers that will be followed up in order to perform a deeper a deeper analysis.The experiment results show its better performance on abnormal detection.