Abstract: There is a lot of knowledge built in to our physical models of stars. But there is even more information in the sum total of all the data ever taken of stars (tens of thousands of pixels of spectral data--brightness vs wavelength data--on many hundreds of thousands of targets). The Cannon (named after Annie Jump Cannon) uses a small amount of physical modeling and a huge amount of data to build very precise, predictive, probabilistic models of stellar spectra. I show that these data-driven models can be used to obtain extremely precise measurements of stellar parameters and detailed chemical abundances, substantially more precise even than the physical models used to generate "ground truth" inputs. Indeed, we believe that The Cannon is delivering more precise chemical abundance measurements than any previous method for the analysis of stellar spectra. These results have implications for studies of extra-Solar planets and the Milky Way. Work in collaboration with Melissa Ness (MPIA), Andrew R. Casey (Cambridge), Anna Y. Q. Ho (Caltech) and Hans-Walter Rix (MPIA).
Bio: David W. Hogg is Professor of Physics and Data Science at New York University. He is a co-founder of NYU's Center for Cosmology and Particle Physics and also its Center for Data Science. He is currently on sabbatical at the Simons Foundation Center for Data Analysis in New York City. He is one of the leads of the multi-university Moore-Sloan Data Science Environments project. Hogg's research is in areas of astrophysics (cosmology, the Milky Way, stars, and exoplanets) where making measurements is extremely difficult. His group produces not just scientific literature but also software tools for astrophysics and other natural sciences.