Willkommen auf den Webseiten der MALEO Gruppe
Die Gruppe Maschinelles Lernen und Optimierung (MALEO) unter der Leitung von Prof. Dr. Heike Trautmann wurde im Oktober 2023 an der Universität Paderborn gegründet. Unsere theoretische und empirische Forschung konzentriert sich auf (vertrauenswürdige) Künstliche Intelligenz, Data Science, (automatisiertes) maschinelles Lernen, automatisierte Algorithmenauswahl und -konfiguration, explorative Landschaftsanalyse, Analyse von randomisierten Suchheuristiken und (mehrkriterielle) evolutionäre Optimierung. In interdisziplinärer Zusammenarbeit befassen wir uns darüber hinaus mit Themen der Computational Social Sciences wie Social Influence Analysis und Disinformation Campaign Detection in Online-Medien. Wir haben vielfältige und umfangreiche nationale sowie internationale Kooperationen mit besonders starken Verbindungen zur Universität Twente, Enschede. Auf europäischer Ebene unterstützen wir die Confederation of Laboratories for Artificial Intelligence Research in Europe (CLAIRE) und das European Research Center for Information Systems (ERCIS).
Für weitere Information verweisen wir auf die offizielle Website unserer Gruppe.
MALEO Mitarbeiter
Aktuelle und vergangene Lehre
- Vorlesung: Angewandte Statistik mit R
Lecturer: Prof. Dr. Heike Trautmann
Die Vorlesung gibt einen Überblick über sowohl deskriptive Statistik und explorative Datenanalyse als auch induktive Statistik und Grundlagen der Wahrscheinlichkeitsrechnung. Der Schwerpunkt liegt neben der Vermittlung von intuitivem Verständnis auf der angewandten Datenanalyse unter Zuhilfenahme der statistischen Programmiersprache R, die im ersten Block der Vorlesung im Rahmen eines Programmierkurses eingeführt wird, entsprechende Vorkenntnisse sind hier nicht erforderlich.
- Vorlesung: Multi-Objective Optimisation
Lecturer: Dr. Jakob Bossek
Optimization problems are ubiquitous, and we all (approximately) solve them in everyday life, such as when finding routes with Google Maps to quickly get from point A to point B or deciding on a checkout lane with the shortest waiting queue (shortest expected waiting time) at the supermarket. However, optimization problems are rarely single-criteria. Instead, they are typically multi-criteria in nature, with the individual objectives usually conflicting with each other. For example, in route planning, the distance traveled may be relevant (shorter is better), and fuel consumption may also be a consideration (lower is better). The shortest route may lead through the city center with many stop-and-go maneuvers at red lights, especially during peak hours. On the other hand, a longer route around the city may consume less fuel. Accordingly, the goal in multi-objective optimization is to find a set of optimal compromise solutions.
This course provides a comprehensive introduction to multi-objective optimization and the associated challenges. In addition to classical general approaches, exact methods for selected combinatorial optimization problems are presented, along with heuristic (nature-inspired) methods. The course also covers heuristic solution approaches for problems with more than three criteria (many-objective optimization).
- Seminar: Machine Learning Assisted Evolutionary Multi- and Many- Objective Optimization
Lecturer: Prof. Dr. Heike Trautmann
This seminar, partly based on an upcoming respective book by Saxena, Deb et al. (2024), focuses on machine learning (ML) assisted evolutionary multi- and many-objective optimization (EMO). EMO algorithms, namely EMOAs, iteratively evolve a set of solutions towards an optimal trade-off surface, i.e. Pareto Front. The availability of multiple solution sets over successive generations makes EMOAs amenable to application of ML for different pursuits, a large proportion of optimization problems in practice is multi-objective in nature.
We will cover the foundations of optimization, including problem and algorithm types and then focus on key studies on ML-based enhancements in the EMO domain, systematically addressing important aspects. Moreover, we will discuss Auto-ML based techniques such as Automated EMOA configuration. The seminar will have two phases, theoretical will be followed by experimental work in groups.
- Seminar: Data Stream Mining
Lecturer: Prof. Dr. Heike Trautmann
This seminar will focus on the characteristics of Data Streams and related Data Stream Mining algorithms. We will mainly concentrate on textual data in the social networks domain and discuss application areas such as Disinformation Campaign Detection or Synthetic Social Network Data Generation. The seminar will have two phases, theoretical will be followed by experimental work in groups.
- Proseminar: Evolutionary Computation
Lecturer: Dr. Jakob Bossek
The key idea of this seminar is to dive into diverse topics in the context of evolutionary computation. Evolutionary algorithms (EAs) are bio-inspired randomized search heuristics that mimic principles from natural evolution to “evolve” good solutions iteratively. Such algorithms particularly shine in the context of finding good solutions for black-box optimization problems (where knowledge about the function at hand is obtainable via queries only) or (multi-objective) NP-hard combinatorial optimisation problems
- Project Group: Next Gen User Interface for optimization software based on Generative AI (GenUIne)
Lecturer: Dr. Moritz Seiler, Prof. Dr. Heike Trautmann
Using generative AI we are going to re-invent the next gen User Interface for optimization software in close cooperation with the company Optano (https://optano.com).
Optimization software (e.g. https://optano.com/en/solutions/) has a wide range of use cases and features, it is hard to learn and complex to use, especially when users are no domain experts or rarely use such tools. For those users the power of optimization remains a locked secret.
This project will demonstrate how latest generative AI, e.g. LLMs, can support users to understand complex software products, get insight in their data and run complex tasks automatically and without deep knowledge of the tool. Users will interact with optimization software using their mother tongue. We will create a great experience for users. We leverage GenAI. We are GenUIne!
- Vorlesung: Unsupervised Learning and Evolutionary Optimisation Using R
- Seminar: Automated Algorithm Selection
Geplante Veranstaltungen
Die folgenden Vorlesungen, Seminare und Projektgruppen sind für die kommenden Semester geplant. Bitte beachten Sie aber, dass die Angaben vorläufig sind und sich noch ändern können!
Geplante Lehre
- Vorlesung: Grundlegende Algorithmen (Bachelor elective, English, 3V + 2Ü)
- Vorlesung: Unsupervised Learning and Evolutionary Computation Using R (Master, English, 3V + 2Ü)
- Projektgruppe: (Master, English) Next Gen User Interface for optimization software based on Generative AI (GenUIne)
- Vorlesung: Multi-Objective Optimisation (Master, English, 3V + 2Ü)
Publikationen
@article{Prager_Trautmann_2024, title={Exploratory Landscape Analysis for Mixed-Variable Problems}, DOI={10.1109/TEVC.2024.3399560}, journal={IEEE Transactions on Evolutionary Computation}, author={Prager, Raphael Patrick and Trautmann, Heike}, year={2024}, pages={1–1} }
@inproceedings{Dietrich_Prager_Doerr_Trautmann_2024, place={Cham}, title={Hybridizing Target- and SHAP-encoded Features for Algorithm Selection in Mixed-variable Black-box Optimization}, booktitle={Parallel Problem Solving from Nature — PPSN XVIII}, publisher={Springer International Publishing}, author={Dietrich, K. and Prager, R. and Doerr, C. and Trautmann, Heike}, editor={Affenzeller, M and Winkler, S and Kononova, A and Trautmann, H and Tušar, T and Machado, P and Baeck, T}, year={2024}, pages={1–14} }
@inproceedings{Seiler_Skvorc_Cenikj_Doerr_Trautmann_2024, place={Cham}, title={Learned Features vs. Classical ELA on Affine BBOB Functions}, booktitle={Parallel Problem Solving from Nature — PPSN XVIII}, publisher={Springer International Publishing}, author={Seiler, Moritz and Skvorc, Urban and Cenikj, G and Doerr, C and Trautmann, Heike}, editor={Affenzeller, M and Winkler, S and Kononova, A and Trautmann, H and Tušar, T and Machado, P and Baeck, T}, year={2024}, pages={1–14} }
@inproceedings{Seiler_Rook_Heins_Preuß_Bossek_Trautmann_2024, title={Using Reinforcement Learning for Per-Instance Algorithm Configuration on the TSP}, DOI={10.1109/ssci52147.2023.10372008}, booktitle={2023 IEEE Symposium Series on Computational Intelligence (SSCI)}, publisher={IEEE}, author={Seiler, Moritz and Rook, Jeroen and Heins, Jonathan and Preuß, Oliver Ludger and Bossek, Jakob and Trautmann, Heike}, year={2024} }
@inbook{Preuß_Rook_Trautmann_2024, place={Cham}, title={On the Potential of Multi-objective Automated Algorithm Configuration on Multi-modal Multi-objective Optimisation Problems}, DOI={10.1007/978-3-031-56852-7_20}, booktitle={Applications of Evolutionary Computation}, publisher={Springer Nature Switzerland}, author={Preuß, Oliver Ludger and Rook, Jeroen and Trautmann, Heike}, year={2024} }
@article{Prager_Trautmann_2023, title={Pflacco: Feature-Based Landscape Analysis of Continuous and Constrained Optimization Problems in Python}, DOI={10.1162/evco_a_00341}, journal={Evolutionary Computation}, author={Prager, Raphael Patrick and Trautmann, Heike}, year={2023}, pages={1–25} }
@article{Bossek_Sudholt_2023, title={Do Additional Target Points Speed Up Evolutionary Algorithms?}, DOI={10.1016/j.tcs.2023.113757}, journal={Theoretical Computer Science}, author={Bossek, Jakob and Sudholt, Dirk}, year={2023}, pages={113757} }
@article{Bossek_Grimme_2023, title={On Single-Objective Sub-Graph-Based Mutation for Solving the Bi-Objective Minimum Spanning Tree Problem}, DOI={10.1162/evco_a_00335}, journal={Evolutionary Computation}, author={Bossek, Jakob and Grimme, Christian}, year={2023}, pages={1–35} }
@article{Heins_Bossek_Pohl_Seiler_Trautmann_Kerschke_2023, title={A study on the effects of normalized TSP features for automated algorithm selection}, volume={940}, DOI={https://doi.org/10.1016/j.tcs.2022.10.019}, journal={Theoretical Computer Science}, author={Heins, Jonathan and Bossek, Jakob and Pohl, Janina and Seiler, Moritz and Trautmann, Heike and Kerschke, Pascal}, year={2023}, pages={123–145} }
@inproceedings{Prager_Dietrich_Schneider_Schäpermeier_Bischl_Kerschke_Trautmann_Mersmann_2023, place={New York, NY, USA}, series={FOGA ’23}, title={Neural Networks as Black-Box Benchmark Functions Optimized for Exploratory Landscape Features}, DOI={10.1145/3594805.3607136}, booktitle={Proceedings of the 17th ACM/SIGEVO Conference on Foundations of Genetic Algorithms}, publisher={Association for Computing Machinery}, author={Prager, Raphael Patrick and Dietrich, Konstantin and Schneider, Lennart and Schäpermeier, Lennart and Bischl, Bernd and Kerschke, Pascal and Trautmann, Heike and Mersmann, Olaf}, year={2023}, pages={129–139}, collection={FOGA ’23} }
Kontakt
Marion Hucke
Machine Learning and Optimisation
Fürstenallee 11
33102 Paderborn
Sprechstunden
Montagnachmittag
Dienstagvormittag
Donnerstagnachmittag