We’re looking for research fellow, postdocs, and doctoral students for projects in
- Bayesian workflows for iterative model building and networks of models (Proj. 7, Aalto University)
-
Evaluating and improving posterior inference for difficult posteriors
(Proj. F9, FCAI/Aalto/Helsinki with Prof. Arto Klami) - Workflows for better priors (Proj. F19, FCAI/Aalto/Helsinki with Prof. Arto Klami)
See the abstracts below.
There are also many other topics in probabilistic modeling, ML, and AI at Aalto University and University of Helsinki
All topics and how to apply at
- research fellows and postdocs: https://www.hiit.fi/fcai-and-project-postdoctoral-researcher-and-research-fellow-positions/
- doctoral students: Helsinki ICT network: Doctoral student positions in computer science--Open positions - Helsinki Doctoral Education Network in ICT
You can ask me for more information
Aalto University and University Helsinki have strong Bayesian/ML/AI community. We contribute to open source software packages like Stan and ArviZ. Aalto pays postdocs well compared to many other countries. We have plenty of travel funds. Finland is a great place for living, with or without family. It is a safe, politically stable and well-organized society, where equality is highly valued and corruption is low. Extensive social security supports people in all situations of life. Occupational and national public healthcare in Finland are great and free. You can manage in work and everyday life well with English (no need to learn Finnish unless you want to). Finland has been ranked as the happiest country in the world in 2018–2021.
–
Topic: Bayesian workflows for iterative model building and networks of models
We formalize and develop theory and diagnostics for iterative Bayesian model building. The practical workflow recommendations and diagnostics guide the modeller through the appropriate steps to ensure safe iterative model building, or indicate when the modeler is likely to be in the danger zone.
–
Topic: Evaluating and improving posterior inference for difficult posteriors
Both MCMC and distributional approximations often struggle to handle complex posteriors, but we lack good tools for understanding how and why. We study diagnostics for identifying the nature of the computational difficulty, e.g. whether the difficulty is caused by narrow funnels or strong curvature. We also develop improved inference algorithms, e.g. via automated and semi-automated transformations.
–
Topic: Workflows for better priors
Bayesian models rely on prior distributions that encode knowledge about the problem, but specifying good priors is often difficult in practice. We are working on multiple fronts on making it easier, with contributions to e.g. prior elicitation, prior diagnostics, prior checking, and specification of priors in predictive spaces.