Feliks Nüske


MPG   MPI DCTS   DMP Group

Computer simulations of processes at the molecular level greatly improve our understanding of the behaviour of biological processes, chemical reactions, and material properties. I'm an applied mathematician working on new algorithms to model and simulate molecular systems, leveraging the power of both data science and physical insight. I'm currently head of the DMP Group at MPI DCTS Magdeburg.

  News

  Research Topics

TT-Network Koopman Methods for Molecular Simulation Molecular simulations lead to huge data sets requiring automated methods to extract equilibrium and kinetic properties. Highlights of our research in this domain include a variational approach, tensor methods, randomized kernel methods.
TT-Network Error Theory for Koopman Modeling Determination of provable error bounds for Koopman models remains a key challenge for data-based dynamical systems research. Some of our recent results cover stochastic systems with ergodicity and kernel methods. Improved results for general ergodic systems were described here.
Greedy Search Data-driven Coarse Graining Learning reduced models that can be simulated more efficiently is an essential step towards scaling up molecular simulations. Our results include a sparse identification method, spectral learning, and error bounds.

  Recent Publications


Full list at Google Scholar.

  Team

  • Feliks Nüske
    		•  me 02/22-:  Max Planck Group Leader, MPI DCTS Magdeburg
    10/23-:  Guest Lecturer, Freie Universität Berlin
    09/19-01/22:  Postdoc, Institute of Mathematics, Paderborn University
    03/17-08/19:  Rice Academy Fellow, CTBP Rice University
    10/12-03/17:  Graduate Student, Freie Universität Berlin
  • Vahid Nateghi
    		•  Vahid 06/22-:  Ph.D. student, MPI DCTS Magdeburg
    09/21-12/21:  Intern, GIPSA-Lab, Universite Grenoble Alpes
    09/18-12/21:  Master's student, DAER, Politecnico di Milano
  • Lei Guo
    		•  Lei 01/23-:  Ph.D. student, MPI DCTS Magdeburg
    04/19-09/22:  M.Sc. Mathematics, Technische Universität Kaiserslautern
  • Minakshi Verma
    		•  Minakshi 09/23-:  Ph.D. student, MPI DCTS Magdeburg
    07/21-05/23:  M.Sc. Mathematics, Indian Institute of Technology Bhubaneswar (IIT Bhubaneswar)
    07/18-05/21:  B.Sc. (Hons.) Mathematics, University of Delhi (DU)

      Main Collaborators

  • Prof Dr Frank Noé, Microsoft Research
    		•  Variational Methods for metastable dynamics
    e.g.     MMS 2013 JCTC 2014 JCP 2016
  • Prof Dr Cecilia Clementi, Freie Universität Berlin
    		•  Data-driven Molecular Coarse Graining
    e.g.     JCP 2018 JCP 2019 ACS Cent Sci 2023
  • Dr Stefan Klus, Heriot-Watt University
    		•  Koopman Operator Methods
    e.g.     PhysD 2020 JCP 2023
  • Prof Dr Karl Worthmann, TU Ilmenau
    		•  Koopman Error Theory
    e.g.     JNLS 2023 PhysD 2024

      Some Presentations

  • Modeling Molecular Kinetics with Koopman Operators and Kernel-based Learning
    • IMSI Workshop Data Sciences for Mesoscale and Macroscale Materials Models, May 2024 (Invited)
    Slides   

  • Kernel Methods for Koopman-based Modeling in Molecular Simulation
    • Chalmers AI4Science Seminar, April 2024 (Invited)
    Slides    Video   

  • Kernel Methods for Koopman-based Modeling
    • GAMM Annual Meeting, March 2024 (MS Talk)
    Slides   

      Contact

       nueske@mpi-magdeburg.mpg.de


       MPI DCTS, Sandtorstr 1, 39106 Magdeburg, GERMANY