Dr. Evelyn Herberg

Postdoctoral Researcher

I have been a postdoctoral researcher at the Scientific Computing and Optimization group at the Interdisciplinary Center for Scientific Computing of Heidelberg University since September 2022.

You can find my CV here.

Research Interests

My research mainly focuses on

  • Mathematical Machine Learning
  • Optimization with partial differential equations
  • Variational discretization
  • Model Order Reduction

Recent teaching

Recent events organized

Currently supervising

  • M.Sc. Thesis of Laurin Ernst:
    Invertible Neural Networks for Parameter Identification in Differential Equations
    M.Sc. Mathematik, Heidelberg University
    Supervision: Roland Herzog and Evelyn Herberg
  • M.Sc. Thesis of Hannah Rickmann:
    Global Convergence Theory for Non-Smooth Newton Methods Applied to Quadratic Programs
    M.Sc. Mathematik, Heidelberg University
    Supervision: Roland Herzog and Evelyn Herberg

Latest publications

  • Mohammed Alshehri, Harbir Antil, Evelyn Herberg and Drew P. Kouri
    An inexact semismooth Newton method with application to adaptive randomized sketching for dynamic optimization
    Finite Elements in Analysis and Design 228, p.104052, 2024
    doi:10.1016/j.finel.2023.104052
    bibtex 
    @ARTICLE{AlshehriAntilHerbergKouri:2024:1,
  AUTHOR = {Alshehri, Mohammed and Antil, Harbir and Herberg, Evelyn and Kouri, Drew P.},
  PUBLISHER = {Elsevier BV},
  DATE = {2024-01},
  DOI = {10.1016/j.finel.2023.104052},
  JOURNALTITLE = {Finite Elements in Analysis and Design},
  PAGES = {104052},
  TITLE = {An inexact semismooth Newton method with application to adaptive randomized sketching for dynamic optimization},
  VOLUME = {228},
}
  • Evelyn Herberg
    Variationelle Diskretisierung für Optimale Steuerung mit Maßkontrollen
    Mitteilungen der Deutschen Mathematiker-Vereinigung 31(3), p.156-159, 2023
    doi:10.1515/dmvm-2023-0053
    bibtex 
    @ARTICLE{Herberg:2023:2,
  AUTHOR = {Herberg, Evelyn},
  PUBLISHER = {Walter de Gruyter},
  DATE = {2023-09},
  DOI = {10.1515/dmvm-2023-0053},
  JOURNALTITLE = {Mitteilungen der Deutschen Mathematiker-Vereinigung},
  NUMBER = {3},
  PAGES = {156--159},
  TITLE = {Variationelle Diskretisierung für Optimale Steuerung mit Maßkontrollen},
  VOLUME = {31},
}
  • Robert J. Baraldi, Evelyn Herberg, Drew P. Kouri and Harbir Antil
    Adaptive randomized sketching for dynamic nonsmooth optimization
    Model Validation and Uncertainty Quantification 3, p.107-116, 2023
    doi:10.1007/978-3-031-37003-8_17
    bibtex 
    @INCOLLECTION{BaraldiHerbergKouriAntil:2023:1,
  AUTHOR = {Baraldi, Robert J. and Herberg, Evelyn and Kouri, Drew P. and Antil, Harbir},
  PUBLISHER = {Springer Nature Switzerland},
  BOOKTITLE = {Model Validation and Uncertainty Quantification},
  DATE = {2023-06},
  DOI = {10.1007/978-3-031-37003-8_17},
  PAGES = {107--116},
  TITLE = {Adaptive randomized sketching for dynamic nonsmooth optimization},
  VOLUME = {3},
}
  • Harbir Antil, Hugo Dı́az and Evelyn Herberg
    An optimal time variable learning framework for deep neural networks
    Annals of Mathematical Sciences and Applications 8(3), p.501-543, 2023
    doi:10.4310/amsa.2023.v8.n3.a4
    bibtex 
    @ARTICLE{AntilDiazHerberg:2023:1,
  AUTHOR = {Antil, Harbir and Dı́az, Hugo and Herberg, Evelyn},
  PUBLISHER = {International Press of Boston},
  DATE = {2023},
  DOI = {10.4310/amsa.2023.v8.n3.a4},
  EPRINT = {2204.08528},
  JOURNALTITLE = {Annals of Mathematical Sciences and Applications},
  NUMBER = {3},
  PAGES = {501--543},
  TITLE = {An optimal time variable learning framework for deep neural networks},
  VOLUME = {8},
}
  • Evelyn Herberg
    Lecture Notes: Neural Network Architectures
    2023
    arXiv:2304.05133
    bibtex 
    @ONLINE{Herberg:2023:1,
  AUTHOR = {Herberg, Evelyn},
  DATE = {2023},
  EPRINT = {2304.05133},
  EPRINTTYPE = {arXiv},
  TITLE = {Lecture Notes: Neural Network Architectures},
}
  • Evelyn Herberg, Roland Herzog and Frederik Köhne
    Time regularization in optimal time variable learning
    Proceedings in Applied Mathematics and Mechanics, 2023
    arXiv:2306.16111 doi:10.1002/pamm.202300299
    bibtex 
    @ARTICLE{HerbergHerzogKoehne:2023:2,
  AUTHOR = {Herberg, Evelyn and Herzog, Roland and Köhne, Frederik},
  DATE = {2023},
  DOI = {10.1002/pamm.202300299},
  EPRINT = {2306.16111},
  EPRINTTYPE = {arXiv},
  JOURNALTITLE = {Proceedings in Applied Mathematics and Mechanics},
  TITLE = {Time regularization in optimal time variable learning},
}
  • Evelyn Herberg, Roland Herzog, Frederik Köhne, Leonie Kreis and Anton Schiela
    Sensitivity-based layer insertion for residual and feedforward neural networks
    2023
    arXiv:2311.15995
    bibtex 
    @ONLINE{HerbergHerzogKoehneKreisSchiela:2023:1,
  AUTHOR = {Herberg, Evelyn and Herzog, Roland and Köhne, Frederik and Kreis, Leonie and Schiela, Anton},
  DATE = {2023},
  EPRINT = {2311.15995},
  EPRINTTYPE = {arXiv},
  TITLE = {Sensitivity-based layer insertion for residual and feedforward neural networks},
}
  • Evelyn Herberg and Michael Hinze
    Variational discretization approach applied to an optimal control problem with bounded measure controls
    Optimization and Control for Partial Differential Equations, p.113-136, 2022
    doi:10.1515/9783110695984-006
    bibtex 
    @INCOLLECTION{HerbergHinze:2022:2,
  AUTHOR = {Herberg, Evelyn and Hinze, Michael},
  PUBLISHER = {De Gruyter},
  BOOKTITLE = {Optimization and Control for Partial Differential Equations},
  DATE = {2022-03},
  DOI = {10.1515/9783110695984-006},
  PAGES = {113--136},
  TITLE = {Variational discretization approach applied to an optimal control problem with bounded measure controls},
}