I did my graduate studies in the Computation Learning & Motor Control Lab under the mentorship of Stefan Schaal. Despite his valiant efforts to get me to engage with robots, I spent most of my days rummaging through books and papers on applied statistics. I particularly liked tinkering with Bayesian nonparametric statistics, and countably infinite models like Gaussian/Dirichlet processes.
My doctoral dissertation applied probabilistic evidence frameworks — and in particular, variational integrals — to understanding issues of inferring (statistical) model complexity. It also yielded a novel way of interpreting the classic Backfitting algorithm through a Bayesian lens.