BIRS Workshop Lecture Videos
Inversion of abelian and non-abelian ray transforms in the presence of statistical noise Monard, Francois
We will discuss two problems associated with ray transforms on simple surfaces: (1) how to reconstruct a function from its noisy geodesic X-ray transform (with applications to X-ray tomography) (2) how to reconstruct a skew-hermitian Higgs field from its noisy scattering data (with applications to Neutron Spin Tomography) For (1), the derivation of new mapping properties for the normal operator I*I, based on a generalization of the transmission condition, allows to prove a Bernsteinâ von Mises theorem, about the statistical reliability of the Maximum A Posteriori as a reconstruction candidate in a Bayesian statistical inversion framework, including a reliable assessment of the credible intervals. For (2), a non-linear problem whose injectivity for the noiseless case was established by Paternainâ Saloâ Uhlmann, the derivation of a new stability estimate allows one to prove a consistency result for the mean of the posterior distribution in the large data sample limit. Numerical illustrations will be presented. Joint works with Gabriel Paternain and Richard Nickl (Cambridge).
Item Citations and Data
Attribution-NonCommercial-NoDerivatives 4.0 International