Galaxy Morphology plays a pivotal role in tracking recent major galaxy mergers, and thus can be used as an indicator in studying how Active Galactic Nuclei (AGN) are triggered and fueled. We apply novel machine learning techniques to analyze images from the state-of-art Subaru Hyper Suprime-Cam Subaru Strategic Program, in order to probe the host galaxies of luminous, heavily obscured AGN over the past 7 billion years. Specifically, my research focuses on removing the central point source from the AGN image, which is dominated by black hole accretion, leaving behind the stellar emission from the host galaxy, whose morphology we then measure. For the first task, I train PSFGAN, a customized Generative Adversarial Network, on simulated and real galaxies that have artificial AGN point sources added. I then train a customized Convolutional Neural Network called GaMorNet on PSFGAN-ed galaxies, adding a Transfer Learning step that trains the final GaMorNet layers on real galaxies with known properties. My work shows that we can predict an AGN host galaxy’s morphological parameters with high accuracy once both networks are optimized. Overall, our method requires far less advance knowledge of galaxy properties than previous machine learning approaches to image analysis. Selected examples w/ a flowchart summarizing the proposed approach are shown in the attachment.