Rianne de Heide
 |
Machine Learning Group, Statistical Science Group |
About me
I am a PhD candidate at CWI under supervision of Peter Grünwald. I work on problems and solutions in machine learning and statistics, with a focus on Bayesian learning, hypothesis testing and sequential learning.
In particular, I am interested in exploring the limitations of existing methods under misspecification of the model or the data collecting process, and in devising new, robust methods for these settings. Recently I became interested in pure exploration problems, which are problems in a bandit setting, dealing with the question how to design systems that efficiently explore their stochastic environment, with the goal of learning something specific about it, for example, best-arm identification.
Currently, I am visiting Alexandra Carpentier at the Otto von Guericke Universität in Magdeburg.
Until August 2020 I was a board member of the European Women in Mathematics – The Netherlands, the Dutch association of women professional mathematicians.
In spring 2019, I visited SequeL (Inria Lille-Nord Europe) on a LUF travel grant, to work with Emilie Kaufmann and Michal Valko.
Research
My research interests include
Bandits, pure exploration problems
(Sequential) hypothesis testing
Group invariance in statistics
Bayesian methods
Learning theory
Foundations of ML, stats and probability theory
Recent Publications
Why optional stopping can be a problem for Bayesians
Rianne de Heide and Peter Grünwald
Psychonomic Bulletin & Review 2020, doi:10.3758/s13423-020-01803-x
Extended technical report: arXiv 1708.08278Optional Stopping with Bayes Factors: A Categorization and Extension of Folklore Results, with an Application to Invariant Situations
Allard Hendriksen, Rianne de Heide and Peter Grünwald
Bayesian Analysis 2020, doi:10.1214/20-BA1234.Safe-Bayesian Generalized Linear Regression
Rianne de Heide, Alisa Kirichenko, Nishant Mehta and Peter Grünwald
AISTATS 2020, PMLR 108:2623-2633.