Fast Modeling and Finding (even more!) Strong Gravitational Lenses with Deep Learning
Saul Perlmutter, Professor
Physics
Applications for Spring 2024 are closed for this project.
Strong gravitational lenses are very rare occurrences and are a powerful tool in studying dark matter and dark energy, two mysterious entities that together account for 95% of the energy in the universe. The strong lensing team works on a range of projects with state-of-the-art computation/machine learning/A.I. techniques. We have found over 3500 strong gravitational lenses in a large imaging survey, the DESI Legacy Surveys (which cover almost the entire extragalactic sky, http://legacysurvey.org/), using deep neural networks. Our Hubble Space Telescope program provides images with exquisite details about these systems. This allows us to construct detailed models of these lenses in order to better understand dark matter and measure the expansion rate of the universe (H0). We have developed a fast GPU-based lens modeling code (the fastest in the world! https://ui.adsabs.harvard.edu/abs/2022ApJ...935...49G/abstract) and will apply it to our systems with Hubble images. We also have a Roman Space Telescope program to build a pipeline to measure H0, using simulated strongly lensed and highly magnified supernovae (we already have a pipeline that can be adapted for this purpose: https://ui.adsabs.harvard.edu/abs/2023ApJ...952...10S/abstract). In addition, we expect to complete our next lens search by the end of 2023. By then we will likely have found 1000 *more* lenses, with the total number over 4500, increasing the number of known lenses by an order of magnitude. Finally DESI (https://www.desi.lbl.gov/) have obtained redshifts for thousands of our lensing systems. This is an unprecedented time to work in strong lensing to address some of the most fundamental questions in cosmology and physics.
Role: The student will focus on one or more of the following:
1. strong lens modeling using Hubble images with GPUs,
2. analyzing DESI spectra for lensing systems, and publishing the results,
3. our next search for new lenses in the imaging data,
4. building a pipeline to search for lenses and lensed supernovae in simulated Roman Space Telescope data
5. using DESI data to find new lenses spectroscopically,
The emphasis this semester will be on 1 and 2.
Qualifications: Proficiency in Python coding a necessity. Experience with numpy/scipy/matplotlib/pandas and TensorFlow/JAX/PyTorch highly desirable, and machine learning a plus.
Preferred but not required (and the order is not so important):
1) experience with parallel (and distributed) computing (we use supercomputers at NERSC, https://www.nersc.gov/, both CPUs and GPUs);
2) experience with training neural nets on GPUs (we train on Google Colab and NERSC);
3) experience with HDF5/PyTables;
4) knowledge in gravitational lensing, or even some experience in modeling gravitational lensing systems;
5) experience with any of the following: modeling galaxy light profiles, image coaddition, and transient detection by image subtraction;
6) spectroscopic observations, data reduction, and fitting for spectral absorption and emission lines;
7) experience with web development.
Day-to-day supervisor for this project: Dr. Xiaosheng Huang, Staff Researcher
Hours: 9-11 hrs
Off-Campus Research Site: Lawrence Berkeley National Lab, Building 50. This project can also be carried out remotely with regular Zoom meetings.
Related website: https://www.youtube.com/watch?v=BdpgOhxEyMs
Related website: https://www.youtube.com/watch?v=BdpgOhxEyMs