Auto Recognition of Solar Radio Bursts Using the C-DCGAN Method

• DOI: 10.3389/fphy.2021.646556
• Publication Date: 2021-09-04T00:10:15Z
• Authors: Weidan Zhang; Fabao Yan; Fuyun Han; Ruopu He; Enze Li; Zhao Wu; Yao Chen
• Keywords: Solar radio

Solar radio bursts can be used to study the properties of solar activities and underlying coronal conditions on basis present understanding their emission mechanisms. With construction observational instruments, around world, a vast volume data has been obtained. Manual classifications these require significant efforts human labor in addition necessary expertise field. Misclassifications are unavoidable due subjective judgments various types strong interference some events. It is therefore timely demanding develop techniques auto-classification or recognition bursts. The latest advances deep learning technology provide an opportunity along this line research. In study, we convolutional generative adversarial network model with conditional information (C-DCGAN) auto-classify bursts, using spectral from Culgoora Observatory (1995, 2015) Learmonth (2001, 2019), metric decametric wavelengths. technique generates pseudo images based available inputs, by modifying layers generator discriminator network. demonstrated that C-DCGAN method reach high-level accuracy auto-recognition And issue caused inadequate numbers samples consequent over-fitting partly resolved.