Chen Gu

Short presentation

Phd student in machine learning

Keywords

• Bayesian inference; Probabilistic modelling

Research

My PhD work is about making Bayesian and probabilistic modelling actually usable for biological and pharmacological systems, especially PK/PD models that are written as ODEs. In these settings the data are often sparse, unbalanced, or heterogeneous, and the models themselves can be only partially or even non-identifiable, which makes standard inference unreliable. I work on understanding and diagnosing the geometry of the posterior (why samplers struggle and what that tells us about the model), on building hierarchical Bayesian models that can borrow strength across individuals or experiments, and on developing clearer ways to quantify joint parameter uncertainty. A parallel part of my work connects modelling with experimental design, i.e. asking what data we should collect so that Bayesian inference becomes more informative while keeping the number of animals or experiments low. Overall, the aim is to lower the practical barriers to using probabilistic programming tools (like Stan) in biomedical and preclinical workflows.