Strong gravitational lensing is a powerful tool in exploring various astrophysical questions, including probing the substructure in dark matter haloes of the lensing galaxies. However one of the main limitations of such analysis is the relatively small number of known lens candidates and confirmed lens systems.
Recent works have shown the potential of CNNs in the task of lens finding — classification of images obtained from the telescopes into lensed and non-lensed systems. Since the number of real lenses is insufficient for training a machine learning algorithm, training datasets heavily rely on simulations. However it has been noticed that CNNs perform worse on lens images obtained with the instrument from the one that simulations were tailored to reproduce (for example, different surveys use different color filters and have different resolution).
The goal of this project is to investigate the prospects of using domain adaptation techniques to bridge the gap between simulated data used for training and real images from different surveys (such as HSC-SSP, HST, DES, JWST, and future missions) and explore which type of lenses has a higher risk of being lost during the automated searches.
Total project length: 175/350 hours.
Intermediate
Python, PyTorch and relevant past experience in Machine Learning.
Please use this link to access the test for this project.
Please DO NOT contact mentors directly by email. Instead, please email ml4-sci@cern.ch with Project Title and include your CV and test results. The relevant mentors will then get in touch with you.