Gaussian Processes for Leisure and Profit

I'll discuss two new applications of Gaussian Processes to cosmology, especially the analysis of quasar spectra. First I'll talk about building a fast model (an emulator) for simulations of the Lyman-alpha forest, providing interpolation of a comparable accuracy to the statistical error from the upcoming DESI survey. Then I'll discuss using a Gaussian Process as a model for the emission spectrum of all SDSS quasars and thus detecting strong absorption signals (damped Lyman-alpha systems) in SDSS, even in noisy spectra, a technique borrowed from Bayesian Machine Learning.

Speaker: Simeon Bird, UC Riverside)