AstroAI Lunch Talks - March 24, 2025 - Zachary Fried
24 Mar 2025 - Joshua Wing
The video can be found here: https://www.youtube.com/watch?v=xwtVEq7JAuo
Speaker: Zachary Fried (MIT)
Title: Automating Chemical Intuition for Molecular Identification and Prediction in Astronomical Observations
Abstract: When manually identifying molecules in radio astronomical observations, scientists often rely on “chemical intuition” to evaluate the plausibility of each molecular candidate. This consideration is important, as the detected molecules are frequently chemically related due to shared environmental processes or common precursor compounds. However, this reliance on human intuition poses a challenge for automation, as replicating such reasoning in code is inherently difficult. To address this, we have developed a new framework that leverages machine learning-based embedding methods to analyze the regions of chemical space occupied by molecules detected in specific interstellar sources. This analysis of chemical space is then integrated into an automated line assignment algorithm. Additionally, by training a model to gain a structured understanding of chemical space, we can generate strong candidate molecules to search for in our astronomical data. In this presentation, I will describe our line-assignment algorithm and demonstrate its application to various interstellar sources.