---
_id: '21199'
abstract:
- lang: eng
text: "As in almost every other branch of science, the major advances in data\r\nscience
and machine learning have also resulted in significant improvements\r\nregarding
the modeling and simulation of nonlinear dynamical systems. It is\r\nnowadays
possible to make accurate medium to long-term predictions of highly\r\ncomplex
systems such as the weather, the dynamics within a nuclear fusion\r\nreactor,
of disease models or the stock market in a very efficient manner. In\r\nmany cases,
predictive methods are advertised to ultimately be useful for\r\ncontrol, as the
control of high-dimensional nonlinear systems is an engineering\r\ngrand challenge
with huge potential in areas such as clean and efficient energy\r\nproduction,
or the development of advanced medical devices. However, the\r\nquestion of how
to use a predictive model for control is often left unanswered\r\ndue to the associated
challenges, namely a significantly higher system\r\ncomplexity, the requirement
of much larger data sets and an increased and often\r\nproblem-specific modeling
effort. To solve these issues, we present a universal\r\nframework (which we call
QuaSiModO:\r\nQuantization-Simulation-Modeling-Optimization) to transform arbitrary\r\npredictive
models into control systems and use them for feedback control. The\r\nadvantages
of our approach are a linear increase in data requirements with\r\nrespect to
the control dimension, performance guarantees that rely exclusively\r\non the
accuracy of the predictive model, and only little prior knowledge\r\nrequirements
in control theory to solve complex control problems. In particular\r\nthe latter
point is of key importance to enable a large number of researchers\r\nand practitioners
to exploit the ever increasing capabilities of predictive\r\nmodels for control
in a straight-forward and systematic fashion."
article_number: '110840'
author:
- first_name: Sebastian
full_name: Peitz, Sebastian
id: '47427'
last_name: Peitz
orcid: 0000-0002-3389-793X
- first_name: Katharina
full_name: Bieker, Katharina
id: '32829'
last_name: Bieker
citation:
ama: Peitz S, Bieker K. On the Universal Transformation of Data-Driven Models to
Control Systems. Automatica. 2023;149. doi:10.1016/j.automatica.2022.110840
apa: Peitz, S., & Bieker, K. (2023). On the Universal Transformation of Data-Driven
Models to Control Systems. Automatica, 149, Article 110840. https://doi.org/10.1016/j.automatica.2022.110840
bibtex: '@article{Peitz_Bieker_2023, title={On the Universal Transformation of Data-Driven
Models to Control Systems}, volume={149}, DOI={10.1016/j.automatica.2022.110840},
number={110840}, journal={Automatica}, publisher={Elsevier}, author={Peitz, Sebastian
and Bieker, Katharina}, year={2023} }'
chicago: Peitz, Sebastian, and Katharina Bieker. “On the Universal Transformation
of Data-Driven Models to Control Systems.” Automatica 149 (2023). https://doi.org/10.1016/j.automatica.2022.110840.
ieee: 'S. Peitz and K. Bieker, “On the Universal Transformation of Data-Driven Models
to Control Systems,” Automatica, vol. 149, Art. no. 110840, 2023, doi:
10.1016/j.automatica.2022.110840.'
mla: Peitz, Sebastian, and Katharina Bieker. “On the Universal Transformation of
Data-Driven Models to Control Systems.” Automatica, vol. 149, 110840, Elsevier,
2023, doi:10.1016/j.automatica.2022.110840.
short: S. Peitz, K. Bieker, Automatica 149 (2023).
date_created: 2021-02-10T07:04:15Z
date_updated: 2023-01-07T12:01:58Z
department:
- _id: '101'
- _id: '655'
doi: 10.1016/j.automatica.2022.110840
intvolume: ' 149'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://www.sciencedirect.com/science/article/pii/S0005109822007075/pdfft?isDTMRedir=true&download=true
oa: '1'
project:
- _id: '52'
name: 'PC2: Computing Resources Provided by the Paderborn Center for Parallel Computing'
publication: Automatica
publication_status: published
publisher: Elsevier
status: public
title: On the Universal Transformation of Data-Driven Models to Control Systems
type: journal_article
user_id: '47427'
volume: 149
year: '2023'
...
---
_id: '27426'
abstract:
- lang: eng
text: "Regularization is used in many different areas of optimization when solutions\r\nare
sought which not only minimize a given function, but also possess a certain\r\ndegree
of regularity. Popular applications are image denoising, sparse\r\nregression
and machine learning. Since the choice of the regularization\r\nparameter is crucial
but often difficult, path-following methods are used to\r\napproximate the entire
regularization path, i.e., the set of all possible\r\nsolutions for all regularization
parameters. Due to their nature, the\r\ndevelopment of these methods requires
structural results about the\r\nregularization path. The goal of this article
is to derive these results for\r\nthe case of a smooth objective function which
is penalized by a piecewise\r\ndifferentiable regularization term. We do this
by treating regularization as a\r\nmultiobjective optimization problem. Our results
suggest that even in this\r\ngeneral case, the regularization path is piecewise
smooth. Moreover, our theory\r\nallows for a classification of the nonsmooth features
that occur in between\r\nsmooth parts. This is demonstrated in two applications,
namely support-vector\r\nmachines and exact penalty methods."
author:
- first_name: Bennet
full_name: Gebken, Bennet
id: '32643'
last_name: Gebken
- first_name: Katharina
full_name: Bieker, Katharina
id: '32829'
last_name: Bieker
- first_name: Sebastian
full_name: Peitz, Sebastian
id: '47427'
last_name: Peitz
orcid: 0000-0002-3389-793X
citation:
ama: Gebken B, Bieker K, Peitz S. On the structure of regularization paths for piecewise
differentiable regularization terms. Journal of Global Optimization. 2023;85(3):709-741.
doi:10.1007/s10898-022-01223-2
apa: Gebken, B., Bieker, K., & Peitz, S. (2023). On the structure of regularization
paths for piecewise differentiable regularization terms. Journal of Global
Optimization, 85(3), 709–741. https://doi.org/10.1007/s10898-022-01223-2
bibtex: '@article{Gebken_Bieker_Peitz_2023, title={On the structure of regularization
paths for piecewise differentiable regularization terms}, volume={85}, DOI={10.1007/s10898-022-01223-2},
number={3}, journal={Journal of Global Optimization}, author={Gebken, Bennet and
Bieker, Katharina and Peitz, Sebastian}, year={2023}, pages={709–741} }'
chicago: 'Gebken, Bennet, Katharina Bieker, and Sebastian Peitz. “On the Structure
of Regularization Paths for Piecewise Differentiable Regularization Terms.” Journal
of Global Optimization 85, no. 3 (2023): 709–41. https://doi.org/10.1007/s10898-022-01223-2.'
ieee: 'B. Gebken, K. Bieker, and S. Peitz, “On the structure of regularization paths
for piecewise differentiable regularization terms,” Journal of Global Optimization,
vol. 85, no. 3, pp. 709–741, 2023, doi: 10.1007/s10898-022-01223-2.'
mla: Gebken, Bennet, et al. “On the Structure of Regularization Paths for Piecewise
Differentiable Regularization Terms.” Journal of Global Optimization, vol.
85, no. 3, 2023, pp. 709–41, doi:10.1007/s10898-022-01223-2.
short: B. Gebken, K. Bieker, S. Peitz, Journal of Global Optimization 85 (2023)
709–741.
date_created: 2021-11-15T09:24:59Z
date_updated: 2023-03-11T17:16:33Z
department:
- _id: '101'
- _id: '655'
doi: 10.1007/s10898-022-01223-2
intvolume: ' 85'
issue: '3'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://link.springer.com/content/pdf/10.1007/s10898-022-01223-2.pdf
oa: '1'
page: 709-741
publication: Journal of Global Optimization
status: public
title: On the structure of regularization paths for piecewise differentiable regularization
terms
type: journal_article
user_id: '47427'
volume: 85
year: '2023'
...
---
_id: '20731'
abstract:
- lang: eng
text: We present a novel algorithm that allows us to gain detailed insight into
the effects of sparsity in linear and nonlinear optimization, which is of great
importance in many scientific areas such as image and signal processing, medical
imaging, compressed sensing, and machine learning (e.g., for the training of neural
networks). Sparsity is an important feature to ensure robustness against noisy
data, but also to find models that are interpretable and easy to analyze due to
the small number of relevant terms. It is common practice to enforce sparsity
by adding the ℓ1-norm as a weighted penalty term. In order to gain a better understanding
and to allow for an informed model selection, we directly solve the corresponding
multiobjective optimization problem (MOP) that arises when we minimize the main
objective and the ℓ1-norm simultaneously. As this MOP is in general non-convex
for nonlinear objectives, the weighting method will fail to provide all optimal
compromises. To avoid this issue, we present a continuation method which is specifically
tailored to MOPs with two objective functions one of which is the ℓ1-norm. Our
method can be seen as a generalization of well-known homotopy methods for linear
regression problems to the nonlinear case. Several numerical examples - including
neural network training - demonstrate our theoretical findings and the additional
insight that can be gained by this multiobjective approach.
article_type: original
author:
- first_name: Katharina
full_name: Bieker, Katharina
id: '32829'
last_name: Bieker
- first_name: Bennet
full_name: Gebken, Bennet
id: '32643'
last_name: Gebken
- first_name: Sebastian
full_name: Peitz, Sebastian
id: '47427'
last_name: Peitz
orcid: 0000-0002-3389-793X
citation:
ama: Bieker K, Gebken B, Peitz S. On the Treatment of Optimization Problems with
L1 Penalty Terms via Multiobjective Continuation. IEEE Transactions on Pattern
Analysis and Machine Intelligence. 2022;44(11):7797-7808. doi:10.1109/TPAMI.2021.3114962
apa: Bieker, K., Gebken, B., & Peitz, S. (2022). On the Treatment of Optimization
Problems with L1 Penalty Terms via Multiobjective Continuation. IEEE Transactions
on Pattern Analysis and Machine Intelligence, 44(11), 7797–7808. https://doi.org/10.1109/TPAMI.2021.3114962
bibtex: '@article{Bieker_Gebken_Peitz_2022, title={On the Treatment of Optimization
Problems with L1 Penalty Terms via Multiobjective Continuation}, volume={44},
DOI={10.1109/TPAMI.2021.3114962},
number={11}, journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
publisher={IEEE}, author={Bieker, Katharina and Gebken, Bennet and Peitz, Sebastian},
year={2022}, pages={7797–7808} }'
chicago: 'Bieker, Katharina, Bennet Gebken, and Sebastian Peitz. “On the Treatment
of Optimization Problems with L1 Penalty Terms via Multiobjective Continuation.”
IEEE Transactions on Pattern Analysis and Machine Intelligence 44, no.
11 (2022): 7797–7808. https://doi.org/10.1109/TPAMI.2021.3114962.'
ieee: 'K. Bieker, B. Gebken, and S. Peitz, “On the Treatment of Optimization Problems
with L1 Penalty Terms via Multiobjective Continuation,” IEEE Transactions on
Pattern Analysis and Machine Intelligence, vol. 44, no. 11, pp. 7797–7808,
2022, doi: 10.1109/TPAMI.2021.3114962.'
mla: Bieker, Katharina, et al. “On the Treatment of Optimization Problems with L1
Penalty Terms via Multiobjective Continuation.” IEEE Transactions on Pattern
Analysis and Machine Intelligence, vol. 44, no. 11, IEEE, 2022, pp. 7797–808,
doi:10.1109/TPAMI.2021.3114962.
short: K. Bieker, B. Gebken, S. Peitz, IEEE Transactions on Pattern Analysis and
Machine Intelligence 44 (2022) 7797–7808.
date_created: 2020-12-15T07:46:36Z
date_updated: 2022-10-21T12:27:16Z
ddc:
- '510'
department:
- _id: '101'
- _id: '530'
- _id: '655'
doi: 10.1109/TPAMI.2021.3114962
file:
- access_level: closed
content_type: application/pdf
creator: speitz
date_created: 2021-09-25T11:59:15Z
date_updated: 2021-09-25T11:59:15Z
file_id: '25040'
file_name: On_the_Treatment_of_Optimization_Problems_with_L1_Penalty_Terms_via_Multiobjective_Continuation.pdf
file_size: 7990831
relation: main_file
success: 1
file_date_updated: 2021-09-25T11:59:15Z
has_accepted_license: '1'
intvolume: ' 44'
issue: '11'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9547772
oa: '1'
page: 7797-7808
publication: IEEE Transactions on Pattern Analysis and Machine Intelligence
publication_status: epub_ahead
publisher: IEEE
status: public
title: On the Treatment of Optimization Problems with L1 Penalty Terms via Multiobjective
Continuation
type: journal_article
user_id: '47427'
volume: 44
year: '2022'
...
---
_id: '16290'
abstract:
- lang: eng
text: The control of complex systems is of critical importance in many branches
of science, engineering, and industry, many of which are governed by nonlinear
partial differential equations. Controlling an unsteady fluid flow is particularly
important, as flow control is a key enabler for technologies in energy (e.g.,
wind, tidal, and combustion), transportation (e.g., planes, trains, and automobiles),
security (e.g., tracking airborne contamination), and health (e.g., artificial
hearts and artificial respiration). However, the high-dimensional, nonlinear,
and multi-scale dynamics make real-time feedback control infeasible. Fortunately,
these high- dimensional systems exhibit dominant, low-dimensional patterns of
activity that can be exploited for effective control in the sense that knowledge
of the entire state of a system is not required. Advances in machine learning
have the potential to revolutionize flow control given its ability to extract
principled, low-rank feature spaces characterizing such complex systems.We present
a novel deep learning modelpredictive control framework that exploits low-rank
features of the flow in order to achieve considerable improvements to control
performance. Instead of predicting the entire fluid state, we use a recurrent
neural network (RNN) to accurately predict the control relevant quantities of
the system, which are then embedded into an MPC framework to construct a feedback
loop. In order to lower the data requirements and to improve the prediction accuracy
and thus the control performance, incoming sensor data are used to update the
RNN online. The results are validated using varying fluid flow examples of increasing
complexity.
article_type: original
author:
- first_name: Katharina
full_name: Bieker, Katharina
id: '32829'
last_name: Bieker
- first_name: Sebastian
full_name: Peitz, Sebastian
id: '47427'
last_name: Peitz
orcid: https://orcid.org/0000-0002-3389-793X
- first_name: Steven L.
full_name: Brunton, Steven L.
last_name: Brunton
- first_name: J. Nathan
full_name: Kutz, J. Nathan
last_name: Kutz
- first_name: Michael
full_name: Dellnitz, Michael
last_name: Dellnitz
citation:
ama: Bieker K, Peitz S, Brunton SL, Kutz JN, Dellnitz M. Deep model predictive flow
control with limited sensor data and online learning. Theoretical and Computational
Fluid Dynamics. 2020;34:577–591. doi:10.1007/s00162-020-00520-4
apa: Bieker, K., Peitz, S., Brunton, S. L., Kutz, J. N., & Dellnitz, M. (2020).
Deep model predictive flow control with limited sensor data and online learning.
Theoretical and Computational Fluid Dynamics, 34, 577–591. https://doi.org/10.1007/s00162-020-00520-4
bibtex: '@article{Bieker_Peitz_Brunton_Kutz_Dellnitz_2020, title={Deep model predictive
flow control with limited sensor data and online learning}, volume={34}, DOI={10.1007/s00162-020-00520-4},
journal={Theoretical and Computational Fluid Dynamics}, author={Bieker, Katharina
and Peitz, Sebastian and Brunton, Steven L. and Kutz, J. Nathan and Dellnitz,
Michael}, year={2020}, pages={577–591} }'
chicago: 'Bieker, Katharina, Sebastian Peitz, Steven L. Brunton, J. Nathan Kutz,
and Michael Dellnitz. “Deep Model Predictive Flow Control with Limited Sensor
Data and Online Learning.” Theoretical and Computational Fluid Dynamics
34 (2020): 577–591. https://doi.org/10.1007/s00162-020-00520-4.'
ieee: K. Bieker, S. Peitz, S. L. Brunton, J. N. Kutz, and M. Dellnitz, “Deep model
predictive flow control with limited sensor data and online learning,” Theoretical
and Computational Fluid Dynamics, vol. 34, pp. 577–591, 2020.
mla: Bieker, Katharina, et al. “Deep Model Predictive Flow Control with Limited
Sensor Data and Online Learning.” Theoretical and Computational Fluid Dynamics,
vol. 34, 2020, pp. 577–591, doi:10.1007/s00162-020-00520-4.
short: K. Bieker, S. Peitz, S.L. Brunton, J.N. Kutz, M. Dellnitz, Theoretical and
Computational Fluid Dynamics 34 (2020) 577–591.
date_created: 2020-03-13T12:40:09Z
date_updated: 2022-01-06T06:52:48Z
department:
- _id: '101'
doi: 10.1007/s00162-020-00520-4
intvolume: ' 34'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://link.springer.com/content/pdf/10.1007/s00162-020-00520-4.pdf
oa: '1'
page: 577–591
project:
- _id: '52'
name: Computing Resources Provided by the Paderborn Center for Parallel Computing
publication: Theoretical and Computational Fluid Dynamics
publication_identifier:
issn:
- 0935-4964
- 1432-2250
publication_status: published
status: public
title: Deep model predictive flow control with limited sensor data and online learning
type: journal_article
user_id: '47427'
volume: 34
year: '2020'
...