Machine Learning of Interstellar Chemical Inventories

The characterization of interstellar chemical inventories provides valuable insight into the and physical processes in astrophysical sources. discovery new molecules becomes increasingly difficult as number viable species grows combinatorially, even when considering only most thermodynamically stable. In this work, we present a novel approach for understanding modeling by combining methodologies from cheminformatics machine learning. Using multidimensional vector representations obtained through unsupervised learning, show that identification candidates astrochemical study can be achieved quantitative measures similarity space, highlighting are similar to those already known medium. Furthermore, simple, supervised learning regressors capable reproducing abundances entire inventories, predict abundance not-yet-seen molecules. As proof-of-concept, have developed applied pipeline inventory well-known dark molecular cloud, Taurus Molecular Cloud 1, one chemically rich regions space date. paper, discuss implications insights explorations provide astrochemistry.