Detecting biosignatures, discovering natural resources, and classifying exoplanets: Adventures in machine learning

Abstract:

Trying to find signs of life, discovering potential sources of natural resources, and the need to characterize exoplanets have always captivated the imagination of scientists and the public alike. Data-driven approaches have been at the center of many recent scientific discoveries and have opened up new frontiers in these areas of inquiry. In this talk, I will take you through some of the most recent and exciting discoveries made using machine learning methods we have developed.

First, we will look at our work on data-driven biosignature detection, where we use network science to characterize planetary atmospheres, and how we combine pyrolysis-gas chromatography-mass spectrometry and machine learning to build an agnostic molecular biosignature. 

Next, we will look at how we used association analysis to predict the locations of as-yet-unknown mineral deposits on Earth and potentially Mars. These advances hold the potential to unlock new avenues of economic growth and sustainable development.

Finally, we will set our sights on exoplanets—celestial bodies orbiting distant stars. The discovery of thousands of exoplanets in recent years has fueled the quest to understand their formation, composition, and potential habitability. We develop machine learning pipelines to better understand and classify exoplanets by embracing the complexity and multidimensionality of exoplanets and their host stars.