Gaussian Processes: the Next Step in Exoplanet Data Analysis

When searching for or characterising exoplanets, we typically need to isolate a deterministic signal from stochastic processes - astrophysical or instrumental "noise" - in time-series data. Gaussian processes (GPs) enable us to construct distributions over random functions, and to infer the properties of "signal" and "noise" in a way that is both flexible and robust. I will give a brief overview of the principles of GPs and show two example applications which are both interesting in their own right, and highlight some specific strengths of the technique. The first is a new re-analysis of the controversial HST/NICMOS transmission spectrum of HD189733b. The second is the measurement of stellar rotation periods from light curves, when the spot distribution evolves over the duration of the dataset.

NB: I could also present another topic: stellar variability studies in Kepler data, based on a new systematics correction which preserves stellar variability. I opted for the GPs because I think it's important to alert the exoplanet community to the potential of this technique, but I'm happy to talk about either.