Talks

In recent decades probing for the subtle indications of new physics in
experimental data has become increasingly difficult. The datasets have gotten
much bigger, the experiments more complex, and the signals ever smaller. Success
stories, like LIGO and Kepler, require a sophisticated combination of statistics
and computation, coupled with an appreciation of both the experimental realities
and the theoretical framework governing the data.

In this talk I will look broadly at data science in physics, and how and why it
has taken an increasingly central role. I'll highlight specifically my current
area of research, radio cosmology: discussing why it is one of the most
challenging areas for data science, and describing my work developing optimal
and efficient statistical methods for turning terabytes of timestreams into
cosmology.

Our physical theories often admit multiple formulations or variants.  Although these variants are generally empirically indistinguishable, they nonetheless appear to represent the world as having different structures.   In this talk, I will discuss several criteria for comparing empirically equivalent theories that may be used to identify (1) when one variant has more structure than another (i.e., when a formulation of a theory has “excess structure”) and (2) when two variants are theoretically equivalent, even though they appear to represent the world differently.  I will then discuss where this leaves the philosopher trying to use our empirically successful theories as a guide to the structure of the world.

I will discuss the cosmology of galileon models with a Minkowski limit and discuss whether they can account for the currently observed cosmological model. The full galileon model predicts the speed of gravitational waves to be different from that of photons. I will discuss this and compare with observations. I will then discuss a subdominant galileon model which is compatible. Finally I will discuss the shape dependence of screening in galileon models, showing that the fifth force is unscreened for planar objects.