---
_id: '60901'
abstract:
- lang: eng
  text: The successful training of deep neural networks is dependent on initialization
    schemes and choice of activation functions. Non-optimally chosen parameter settings
    lead to the known problem of exploding or vanishing gradients. This issue occurs
    when gradient descent and backpropagation are applied. For this setting the Ensemble
    Kalman Filter (EnKF) can be used as an alternative optimizer when training neural
    networks. The EnKF does not require the explicit calculation of gradients or adjoints
    and we show this resolves the exploding and vanishing gradient problem. We analyze
    different parameter initializations, propose a dynamic change in ensembles and
    compare results to established methods.
author:
- first_name: Alper
  full_name: Yegenoglu, Alper
  id: '117951'
  last_name: Yegenoglu
  orcid: 0000-0001-8869-215X
- first_name: Kai
  full_name: Krajsek, Kai
  last_name: Krajsek
- first_name: Sandra Diaz
  full_name: Pier, Sandra Diaz
  last_name: Pier
- first_name: Michael
  full_name: Herty, Michael
  last_name: Herty
citation:
  ama: 'Yegenoglu A, Krajsek K, Pier SD, Herty M. Ensemble Kalman Filter Optimizing
    Deep Neural Networks: An Alternative Approach to Non-performing Gradient Descent.
    In: <i>Lecture Notes in Computer Science</i>. Springer International Publishing;
    2021. doi:<a href="https://doi.org/10.1007/978-3-030-64580-9_7">10.1007/978-3-030-64580-9_7</a>'
  apa: 'Yegenoglu, A., Krajsek, K., Pier, S. D., &#38; Herty, M. (2021). Ensemble
    Kalman Filter Optimizing Deep Neural Networks: An Alternative Approach to Non-performing
    Gradient Descent. In <i>Lecture Notes in Computer Science</i>. Springer International
    Publishing. <a href="https://doi.org/10.1007/978-3-030-64580-9_7">https://doi.org/10.1007/978-3-030-64580-9_7</a>'
  bibtex: '@inbook{Yegenoglu_Krajsek_Pier_Herty_2021, place={Cham}, title={Ensemble
    Kalman Filter Optimizing Deep Neural Networks: An Alternative Approach to Non-performing
    Gradient Descent}, DOI={<a href="https://doi.org/10.1007/978-3-030-64580-9_7">10.1007/978-3-030-64580-9_7</a>},
    booktitle={Lecture Notes in Computer Science}, publisher={Springer International
    Publishing}, author={Yegenoglu, Alper and Krajsek, Kai and Pier, Sandra Diaz and
    Herty, Michael}, year={2021} }'
  chicago: 'Yegenoglu, Alper, Kai Krajsek, Sandra Diaz Pier, and Michael Herty. “Ensemble
    Kalman Filter Optimizing Deep Neural Networks: An Alternative Approach to Non-Performing
    Gradient Descent.” In <i>Lecture Notes in Computer Science</i>. Cham: Springer
    International Publishing, 2021. <a href="https://doi.org/10.1007/978-3-030-64580-9_7">https://doi.org/10.1007/978-3-030-64580-9_7</a>.'
  ieee: 'A. Yegenoglu, K. Krajsek, S. D. Pier, and M. Herty, “Ensemble Kalman Filter
    Optimizing Deep Neural Networks: An Alternative Approach to Non-performing Gradient
    Descent,” in <i>Lecture Notes in Computer Science</i>, Cham: Springer International
    Publishing, 2021.'
  mla: 'Yegenoglu, Alper, et al. “Ensemble Kalman Filter Optimizing Deep Neural Networks:
    An Alternative Approach to Non-Performing Gradient Descent.” <i>Lecture Notes
    in Computer Science</i>, Springer International Publishing, 2021, doi:<a href="https://doi.org/10.1007/978-3-030-64580-9_7">10.1007/978-3-030-64580-9_7</a>.'
  short: 'A. Yegenoglu, K. Krajsek, S.D. Pier, M. Herty, in: Lecture Notes in Computer
    Science, Springer International Publishing, Cham, 2021.'
date_created: 2025-08-06T15:02:38Z
date_updated: 2025-08-08T11:36:59Z
doi: 10.1007/978-3-030-64580-9_7
language:
- iso: eng
place: Cham
publication: Lecture Notes in Computer Science
publication_identifier:
  isbn:
  - '9783030645793'
  - '9783030645809'
  issn:
  - 0302-9743
  - 1611-3349
publication_status: published
publisher: Springer International Publishing
status: public
title: 'Ensemble Kalman Filter Optimizing Deep Neural Networks: An Alternative Approach
  to Non-performing Gradient Descent'
type: book_chapter
user_id: '117951'
year: '2021'
...