Yu Wei Lin(NTHU)
Far-Infrared (FIR) observation is important for studying galaxies and star-formation(SF). The FIR plays an important role in estimating the star-formation rate (SFR) because the dust heated by stellar light re-emits thermal radiation in the infrared. Especially, FIR data points around the dust peak are critical to obtain more accurate dust properties (e.g., temperature, luminosity). However, available data is limited and there are no available FIR telescopes covering 100~300 um ranges, where the FIR dust peak of galaxies normally appear. For example, a large fraction of the IR galaxies detected in the NEP survey by AKARI space telescope do not have FIR data. In this work, we take advantage of all avalilable samples having FIR measurements by Herschel/SPIRE as training samples. We tested two machine learning (ML) algorithms, e.g., support vector machine (SVM) and XGBoost, to estimate the FIR fluxes of the SPIRE bands.