|Speaker||Talk Date||Talk Series|
|Suzanne Aigrain (Oxford)||14 June, 2012||Institute of Astronomy Colloquia|
Transiting exoplanets are extremely valuable because the transit geometry enables very detailed physical characterisation of the planets and host stars. However, the planetary signals of interest are often dwarfed by the host star's intrinsic variability and by instrumental effects, both of which are stochastic and usually have a "red" power spectrum. In my talk I will introduce novel statistical methods imported from the machine learning literature, which enable us to learn the properties of the noise from the data. I will describe two applications: Hubble Space Telescope transmission spectroscopy of hot Jupiters and characterisation of stellar variability in Kepler data.