Description

Leader of the group Scientific Computing: Daan Crommelin.

The SC groups develops efficient mathematical methods to simulate and predict real-world phenomena with inherent uncertainties. Such uncertainties arise from e.g. uncertain model parameters, chaotic dynamics or intrinsic randomness, and can have major impact on model outputs and predictions. Our work is targeted in particular at applications in climate, energy, and finance. In these vital areas, the ability to assess uncertainties and their impact on model predictions is of paramount importance. Expertise in the SC group includes uncertainty quantification, data assimilation, stochastic multiscale modeling and risk assessment. The availability of data to inform and improve simulations and predictions, for example through learning and data-driven modeling, plays an important role in our research.

 

More

 

 

 

 

 

Vacancies

No vacancies currently.

News

Current events

SC Seminar Benjamin Sanderse (CWI)

  • 2021-03-11T15:30:00+01:00
  • 2021-03-11T16:30:00+01:00
March 11 Thursday

Start: 2021-03-11 15:30:00+01:00 End: 2021-03-11 16:30:00+01:00

online

Seminar on Machine Learning and Uncertainty Quantification for Scientific Computing

Online https://cwi-nl.zoom.us/j/9201774084?pwd=OFBqM1dUenFreGdPUWEwZFYvMlJ6UT09

 

Multi-Level Neural Networks for PDEs with Uncertain Parameters

Benjamin Sanderse (CWI)

A novel multi-level method for partial differential equations with uncertain parameters is proposed. The principle behind the method is that the error between grid levels in multi-level methods has a spatial structure that is by good approximation independent of the actual grid level. Our method learns this structure by employing a sequence of convolutional neural networks, that are well-suited to automatically detect local error features as latent quantities of the solution. Furthermore, by using the concept of transfer learning, the information of coarse grid levels is reused on fine grid levels in order to minimize the required number of samples on fine levels. The method outperforms state-of-the-art multi-level methods, especially in the case when complex PDEs (such as single-phase and free-surface flow problems) are concerned, or when high accuracy is required.

Members

Associated Members

Publications

Current projects with external funding

  • Valuation Adjustments for Improved Risk Management (ABC-EU-XVA)
  • Physics based ICT: The digital twin in pipelines (DP-Trans)
  • Discretize first, reduce next: a new paradigm to closure for fluid flow simulation (None)
  • Unravelling Neural Networks with Structure-Preserving Computing (Unravelling Neural Networks)
  • Verified Exascale Computing for Multiscale Applications (VECMA)

Related partners

  • Max Planck Institute for Informatics
  • Bull Sas
  • CBK Sci Con Ltd
  • Bayerische Akademie der Wissenschaften
  • MeitY
  • Instytut Chemii Bioorganicznej Polskiej Akademii Nauk
  • Technische Universiteit Eindhoven
  • Brunel University London
  • University College London
  • Universiteit Leiden
  • Universiteit van Amsterdam