Calendar of Events

Speaker: Kartheik Iyer (University of Toronto)

Describing the star formation histories (SFHs) of galaxies has traditionally been done using functional forms. However, these functional forms often fall short when it comes to capturing the complexity found in the SFHs of galaxies in simulations and nearby galaxies. I will introduce the general problem of describing complex functional forms with observational and systematic uncertainties, and highlight a proposed solution using Gaussian processes (GPs) - a generalization of the Gaussian distribution to the space of functions. I'll describe GPs in more detail, and run through the ingredients needed to build one from scratch. I'll also describe the kernel (or covariance function) - a key ingredient in constructing a GP and a way to encode physical information into the GP priors. These methods can be generalised to a large class of problems in astrophysics and beyond.