Machine Learning for Detection of Multi-Planet Signals in Radial Velocity Surveys
Shvetha Chynoweth
Department of Physics & Astronomy
Faculty Supervisor: John Brewer
Long running radial velocity surveys of exoplanets have to deal with large amounts of data which then relies heavily on manual inspection to determine promising candidates for further investigation. This manual inspection process extends timelines as it relies on continuous human intervention, which struggles to keep up with the rate of data collection. Our aim here is to adopt a standardized and automated algorithmic approach for the identification of promising signals within our dataset. We explored existing machine learning (ML) models that are very successful, but do not align with our objectives since they are primarily designed for transit-based retrieval. The radial velocity-based ML space has definitely seen improvements but is still not entirely free from the requirement of constant human intervention and contains a huge number of unexplored avenues. I'll be presenting here the current state of our machine learning model which is primarily designed to meet this objective for an automated approach that can help identify multi-planet signals as a part of the data retrieval pipeline with EXPRES, while highlighting both their successes and challenges