Juho Piironen

Juho Piironen · Thu, 04.05.2017

I attended the 20th International Conference on Artificial Intelligence and Statistics (AISTATS) which took place in Fort Lauderdale, Florida, in April 20-22 2017.  During the trip I also visited Columbia University, New York, to meet our collaborators (Prof. Andrew Gelman and his group). The purpose of the meeting was to assist our work on the software implementations for the methods we study, in particular, help making our methods available to the users of Stan, which is currently the most widely used probabilistic programming language.

In the conference I presented our paper 'On the Hyperprior Choice for the Global Shrinkage Parameter in the Horseshoe Prior' of which I was the first author (co-author being my supervisor Prof. Aki Vehtari) and which will be part of my dissertation. In addition to the poster, the paper was selected to be presented as an oral talk. The paper discusses Bayesian statistical inference and prior specification in high-dimensional problems, where the goal is to model the target variable of interest with a large number of covariates. Inference in these sorts of problems is the core of my doctoral thesis, and the paper was also also on-topic for the conference which is one of the top conferences in the field.

The conference arrangements and the venue were very nice, and I met a lot of new people. Overall I enjoyed the conference program, networking with the other participants and discussing their work, as well as the nice weather in Florida. I would like to thank HICT for supporting trips like this, which are definitely valuable experience for any doctoral student's scientific career.