Description
Multiple research domains ranging from physics to bio-medical research have started to incorporate simulators to answer scientific questions that have eluded us thus far due to lack of computational power and the modelling capabilities. Simulator-based inference is a computational tool that has emerged from the intersection of statistics and machine learning and that is rapidly growing. It enables to calibrate mechanistic models using computer simulations as the main tool. We will develop new efficient probabilistic surrogate models for the simulator-based inference problem. The surrogate modelling can be used in multiple ways in simulation-based inference, as it can be used to replace virtually any or all components of the stochastic simulation system. We will develop theoretical guarantees for our new methods. We will investigate whether our new methods for quantifying the difference between synthetic simulation data and observational real-world data will outperform workflow-based human-in-the-loop strategies.
Specific project requirements
-
Master’s degree in statistics, computer science, data science, mathematics, or a related quantitative subject with proven competence in statistics and/or machine learning.
-
Documented experience in scientific programming (especially Python) will be an advantage.
Supervisors
-
Professor Jukka Corander, jukka.corander@medisin.uio.no contact person for inquiries about the position