3:45–4:45 pm
Gordon Center for Integrative Science GCIS W301/W303 929 E. 57th Street Chicago, IL 60637
Professor Eric Jonas in the Computer Science Department at UChicago, will be giving a talk as part of the James Franck Institute's Emerging Frontiers seminar series.
Abstract: Spectroscopic measurements of molecules and mixtures can be viewed as inverse problems, where we often have good knowledge of the forward (measurement) process, and wish to recover unknowns (such as mixture concentrations or molecular structure). However, the measurement process is highly nonlinear (quantum mechanical) and the answers we seek have considerable structure and constraints (such as points on a simplex or discrete connected graphs). Focusing on connected graphs, I will present our recent developments of machine-learning structured prediction techniques that enable accurate prediction of the spectrum from the structure (the forward problem) and the recovery of the molecular structure from the observed spectrum (the inverse problem). I will demonstrate the success of these techniques on experimental mass spectrometry (MS) and nuclear magnetic resonance (NMR) spectroscopy data, and if time permits discuss future directions with optical and rotational spectra.
Host: Suri Vaikuntanathan, 773-702-7256 or via email at svaikunt@uchicago.edu. Persons who may need assistance please contact Brenda Thomas at 773-702-7156 or by email at bthomas@uchicago.edu