Thousands of new worlds beyond our own solar system have been discovered, revealing a hugely diverse exoplanetary architecture.
Exoplanets form in evolving protoplanetary accretion discs. The conditions in these discs decide the final mass and ultimate orbital configuration of their exoplanetary systems, causing diversity in the exoplanet architecture.
As exoplanets form, they leave behind signatures of their formation that can be detected in interferometric mm observations, such as rings and spirals.
In order to try and measure the mass of these forming in planets inside their nascent discs, we typically perform around 100 dusty fluid simulations for each observed system, and try to get the mass this way. However, this is incredibly inefficient, inaccurate, and profoundly limits the regions of parameter space we can explore.
At UGA, I am building a research group that will move past this outdated model by harnessing the power of machine learning and information extraction. We are developing neural network techniques that are widely applicable, user-friendly, and around 10,000 times more computationally efficient than current approaches to determining exoplanet mass in forming systems.