Project: Deep Removal of Active Galactic Nuclei
Description
Almost all massive galaxies harbor a central supermassive black hole (SMBH). As these SMBHs grow by accreting surrounding gas and dust, they liberate tremendous amounts of energy, becoming visible as Active Galactic Nuclei (AGN). While the AGN phase is incredibly important for understanding galaxy evolution, its intense luminosity presents a major observational challenge: the bright central point source often outshines and obscures the structural details of the host galaxy itself.
Currently, researchers rely on traditional AGN-host galaxy decomposition techniques using parametric modeling tools like GALFIT. However, these methods are computationally expensive, require significant manual fine-tuning, and struggle to scale with large, modern datasets.
This project aims to develop a novel, data-driven alternative: Deep Removal of AGN (DRAGN). DRAGN is a deep learning framework designed to cleanly remove AGN contributions from astronomical images.
Your primary goal will be to develop a generative framework based on diffusion models to perform image-to-image (i2i) translation. The model will map AGN-contaminated images to AGN-free galaxy images.
To train and validate the DRAGN framework, you will utilize a robust, custom-built dataset of over 500,000 simulated galaxy-AGN pairs. Crucially, these simulations are designed to perfectly mimic the high-resolution observations obtained from the James Webb Space Telescope (JWST) NIRCam F150W filter, ensuring the model is primed for cutting-edge astronomical data.
The figure below shows examples of performance on mock data.
Details
- Supervisor
-
Vlado Menkovski
- Secondary supervisor
-
LWLingyu Wang
- Interested?
- Get in contact