Numerical Challenges in Computational Astrochemistry
Speaker: Tommaso Grassi (Max Planck for Extraterrestrial Physics)
Abstract:
Astrophysical numerical models encounter substantial computational challenges when integrating complex, time-dependent chemistry with physical processes. To address these issues, I will present the use of autoencoders for the dimensionality reduction of chemical networks, enabling efficient solutions with standard ODE solvers while preserving key network features. Additionally, I will discuss the application of interpretable machine learning techniques to connect synthetic spectra with model parameters, facilitating the assessment of information retention in observational data.