I will be working with Dr. Tiziana Di Matto and Dr. Rupert Croft at Carnegie Mellon to develop applications of super-resolution techniques to black hole simulations, using Generative Adversarial Networks based on this previous work with Yin Li, Yueying Ni, Simeon Bird, and Yue Feng.
A shorter summary of their previous work can be found here: https://www.cmu.edu/cosmology/news/articles/2021-05-05_supersims.html
Image credit: Yin Li, Yueying Ni , Rupert A. C. Croft, Tiziana Di Matteo, Simeon Bird, and Yu Feng
Star formation with Dr. Stella offner and Dr. Dávid Guszejnov
Summer 2020 - Origins of Mass Segregation in Stellar Clusters within the STARFORGE Simulations
Stellar clusters that form within Giant Molecular Clouds (GMCs) are mass segregated. This is either because massive stars are born at the center of the clusters, or because they migrate to the center due to gravitational interactions. We analyzed a run from the STARFORGE simulation suite to identify what role the above processes play in determining mass segregation. We found that clusters begin as inversely- or non-mass segregated but contain separate mass-segregated star formation sites. As the cluster evolves, these substructures merge, and the overall cluster becomes more mass segregated due to gravitational interactions.
See the full article here: https://iopscience.iop.org/article/10.3847/2515-5172/abba78
This work was presented at the 237th American Astronomical Society Meeting: https://baas.aas.org/pub/2021n1i131p07/release/1?readingCollection=27a4f9bb
Summer 2021 - Machine Learning Driven Analysis of Core Evolution in Giant Molecular Clouds
Poster from informal department symposium: https://docs.google.com/presentation/d/1y80rXFFuxVV4KuW1rxsuRCkOK8DATP3-QIA46nf6gmg/edit?usp=sharing
I plan to present this work at the 239th American Astronomy Society Meeting
Supernovae with Dr. Dan milisavjevic
Academic Year 2017-18, 2018-19, 2019-20, 2020-21 - Time Domain Astrophysics
Over the course of 2018 and 2019, I designed and created Linux and Python based pipelines to process massive datasets from the Rochester Astronomy Bright Supernova database using statistical analyses. This included scraping the Rochester Astronomy Bright Supernova database to check for supernova pairs that are spatially within three arcseconds of each other and significantly temporally separated. I also started the development of a comprehensive supernova coordinate database from 45,000 Bright Supernova database images, built from a pipeline integrating Astrometry and Source Extractor to identify reliable supernova coordinates based on rough pre-existing coordinates.
In 2020 and 2021, I aided in the training of the Recommender Engine for Intelligent Transient Tracking by creating a Gaussian process regression-based methodology for the augmentation of the Zwicky Transient Facility supernova light curve data set. I also helped develop the accuracy of REFITT and the ease of its workflow by "vetting" the produced lists of supernovae for viable candidates for follow-up.
Galaxy evolution with Dr. Brian o'Shea and Dr. Benoit Côté
Summer 2019 - GAMMA-EM: Emulating of Galactic Chemical Evolution Models to Explore the Galactic Origins of the Elements
The elements on the surface of stars carry a permanent snapshot of the star formation and chemical evolution history of a galaxy. When modern models of galactic chemical evolution are compared to these snapshots, it should be possible to discern the chemical evolution process in a galaxy. However, current models are too time-intensive to evaluate with proper statistical methods, which require many iterations of the model within its parameter space to produce a probability distribution of the starting parameters that best fit the observations. As one proposed solution, we aim to emulate the Galaxy Assembly with Merger Trees for Modeling Abundances (GAMMA) model through the use of Gaussian process regression, then compare the results to newly available observational data with Markov Chain Monte Carlo methods. By training a Gaussian process-based emulator with numerous training GAMMA samples generated from a sparsely sampled set of input parameters, we greatly reduce the computational time required to produce chemical evolution predictions from GAMMA. Given this, we expect to use this emulator model (GAMMA-EM) in conjunction with Markov Chain Monte Carlo to obtain a set of GAMMA input parameters that produce the best model fit to newly available observational data. This will likely improve our current understanding of the chemical evolution process in our galaxy and many others.
Link to Github repo with full project and code: https://github.com/cmarkey/GAMMA-EM
This work was presented at the 235th American Astronomical Society Meeting: https://ui.adsabs.harvard.edu/abs/2020AAS...23520708M/abstract. Poster: https://docs.google.com/presentation/d/1BzsKLIlxa6Dy98jYU0jISjWWFdp9Qok5K82YffC-N6c/edit?usp=sharing