Joseph Richards
UC Berkeley
ABSTRACT
Next-generation photometric surveys need sophisticated
statistical tools that accurately classify variable sources for
optimal allocation of follow-up resources. I will discuss some of the
challenges in building such a classifier, including estimation of
class-predictive features from photometric light curves and
classification modeling using flexible, robust methods. I describe my
work on two classification problems: supernova typing using
simulations from the DES Supernova Classification Challenge and
variable star classification using light curves from the OGLE and
Hipparcos surveys. I will detail my use of cutting-edge statistical
methodology such as diffusion maps, random forests and structured
classification for these problems.