On Learning from Chemical Spectra as Sets of Localized Events

Abstract: This talk presents ongoing work on a representation-learning framework for chemical spectra that models spectra as localized structure rather than conventional representations. The work investigates whether incorporating additional information about spectral morphology can improve reconstruction, interpretability, and downstream learning while preserving physically meaningful structure.

Bio: Reinaldo Mock is a 3rd-year ML Ph.D. student at Georgia Tech. His background is in biomedical engineering and mathematics. His research focuses on developing interpretable, field-ready ML methods for scientific sensing and decision-making. He is advised by Professor Jing Li and GTRI researchers Roman Aranda and Eric Pooser. His current interests include physics-informed machine learning, representation learning, and uncertainty quantification.