Bo Han Chen(Department of Physics, National Tsing Hua University);Tomotsugu Goto(Institute of Astronomy, National Tsing Hua University);Seong Jin Kim(Institute of Astronomy, National Tsing Hua University);Ting Wen Wang(Institute of Astronomy, National Tsing Hua University)
For the purpose of understanding cosmic accretion history of supermassive blackholes(SMBHs), a reliable way on recognizing active galactic nucleus(AGNs) from star-forming galaxies(SFGs) plays an important role. However, AGNs are usually obscured in UV and soft X-ray observations, and are also hard to be distinguished in mid-infrared(MIR) because SFGs have strong polycyclic aromatic hydrocarbon emission in MIR too. Thus, a reliable solutionon recognizing AGNs still remains unsolved. In this work, we provide an novel AGN recognition method based on Deep Neural Network,which significantly improve not only recovering rate on AGNs but also accuracy on classifying AGNs and SFGs. The Deep Neural Network takes 44 band magnitudes and errors as input, and gives the AGN/SFG classification result as output. In conclusion, the recognition accuracy is around 80.29% - 85.15%, with AGN recovering rate is around 85.83% - 88.04%.