Student: Dustin Swarm, Graduate Student in Physics and Astronomy, University of Iowa
Faculty Advisor: Casey DeRoo
Employing Machine Learning Algorithms to Search for Unique X-ray Sources
Astronomical surveys provide more data than astronomers can process through traditional means. This requires new astronomical methods to sift through large datasets and prioritize scientific investigation. Machine learning techniques are a key method being leveraged by astronomers to process these large datasets. Under the supervision of my advisor, Prof. Casey DeRoo, I will develop unsupervised machine learning algorithms (e.g. Random Forest, Kohonen map) to investigate the Chandra Source Catalog (CSC), a database of approximately 315,000 X-ray sources observed by Chandra over the past two decades. The unsupervised algorithm will categorize the X-ray sources in the CSC by grouping together sources with similar physical properties. We will compare the results of many runs of the algorithm to find sources that are consistently categorized by themselves, showing they differ from expected source categories in the CSC. My future work will focus on making follow up observations of these uncharacterized sources to analyze what makes them unique. This work leverages the substantial data archive built by NASA’s Chandra X-ray Observatory to extend scientific discovery beyond individual observations. Looking for outlier sources within the CSC increases the odds of discovering sources that offer new insight into astrophysical phenomena or are in a relatively short-lived evolutionary state. Furthermore, we will develop a blueprint for addressing the problem of data saturation in a manner that is easily applicable to other large datasets, such as those of JWST and WFIRST.