Oral Presentation

Machine learning based approach for merger classification in HSC-SSP and quantitative investigations on role of environment in merger activity

Author(s): Kiyoaki Christopher Omori (Nagoya U.), Tsutomu T. Takeuchi (Nagoya U., ISM), Connor Bottrell (Kavli IPMU), Mike Walmsley (U. of Manchester), Hassen Yusuf (Kavli IPMU), Andy Goulding (Princeton U.), John Silverman (Kavli IPMU), Xuheng Ding (Kavli IPMU)

Presenter: Kiyoaki Christopher Omori (Nagoya University)

Galaxy interactions and mergers are an important process to consider when discussing galaxy evolution. Galaxy interactions can drive various processes pertaining to galaxy evolution, such as inducing star formation and accelerating the accretion of gas onto supermassive black holes and the subsequent ignition of active galactic nuclei (AGN). However, the relative impact of galaxy mergers in the context of galaxy evolution is still unclear, and further investigations are required to improve our understanding. For this work, we prepare a merger sample from Subaru HSC-SSP using a transfer learning-based approach in fine-tuning. Transfer learning is an approach where the weights of a pre-trained model are re-used for a new problem, allowing for improved performances even with smaller training datasets. Fine-tuning uses the pre-trained model as a base, but with new output layers trained for the new problem, which in the case of this work is merger classification. We fine-tune Zoobot (Walmsley et al. 2022), a pre-trained model trained on galaxy images and labels of DeCALS, with HSC-SSP galaxy image and labels. We will discuss the performance of our classifier, and further discuss our findings of the role of environment in galaxy merger incidence.