The DFG-funded project "Robust sampling for Bayesian neural networks” (joint project together with colleagues from TU Freiberg and University of Passau) aims to develop efficient numerical methods for the simulation and evaluation of Bayesian neural networks. This enables uncertainty quantification in deep learning.
The group ‘Numerical analysis of stochastic and deterministic partial differential equations’ at Freie Universität Berlin (https://www.mi.fu-berlin.de/math/groups/naspde) focuses is on applied and computational mathematics, in particular optimization, inverse problems and uncertainty quantification.

Job description:
The project aims at efficient procedures to evaluate the posterior distribution, in particular, in scenarios of deep neural networks with very informative data. The successful candidate will conduct research in the field numerical analysis, applied stochastics, high-dimensional Bayesian statistics, and/or machine learning. The third-party funded research project provides an opportunity to do a doctorate.

Requirements:
A completed scientific university degree (master’s degree) in mathematics or a closely related field is required.

Desirable:
Solid mathematics education (focus on numerical analysis, functional analysis or stochastic analysis), programming experience.

For further information, please contact Prof. Dr. Claudia Schillings (c.schillings@fu-berlin.de / +4930 83875803).