Maximilian Stubbemann

I have left the Knowledge and Data Engineering Group on 31.10.2023. I am now working at the Information Systems and Machine Learning Lab of the University of Hildesheim. My new homepage is available here.
Email: stubbemann@cs.uni-kassel.de

Publications

• 1.

  Stubbemann, M., Hanika, T., Schneider, F.M.: Intrinsic Dimension for Large-Scale Geometric Learning. Transactions on Machine Learning Research. (2023).

  URLBibTeXEndNote

  @article{stubbemann2022intrinsic,
  author = {Stubbemann, Maximilian and Hanika, Tom and Schneider, Friedrich Martin},
  journal = {Transactions on Machine Learning Research},
  keywords = {itegpub},
  title = {Intrinsic Dimension for Large-Scale Geometric Learning},
  year = 2023
  }
  

  %0 Journal Article
  %1 stubbemann2022intrinsic
  %A Stubbemann, Maximilian
  %A Hanika, Tom
  %A Schneider, Friedrich Martin
  %D 2023
  %J Transactions on Machine Learning Research
  %T Intrinsic Dimension for Large-Scale Geometric Learning
  %U https://openreview.net/forum?id=85BfDdYMBY
  

• 1.

  Stubbemann, M., Hille, T., Hanika, T.: Selecting Features by their Resilience to the Curse of Dimensionality. (2023).

  BibTeXEndNote

  @article{stubbemann2023selecting,
  author = {Stubbemann, Maximilian and Hille, Tobias and Hanika, Tom},
  keywords = {itegpub},
  title = {Selecting Features by their Resilience to the Curse of Dimensionality},
  year = 2023
  }
  

  %0 Journal Article
  %1 stubbemann2023selecting
  %A Stubbemann, Maximilian
  %A Hille, Tobias
  %A Hanika, Tom
  %D 2023
  %T Selecting Features by their Resilience to the Curse of Dimensionality
  

• 1.

  Stubbemann, M., Stumme, G.: The Mont Blanc of Twitter: Identifying Hierarchies of Outstanding Peaks in Social Networks. In: European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML-PKDD 2023. pp. 177–192. Springer (2023). https://doi.org/10.1007/978-3-031-43418-1\_11.

  URLBibTeXEndNote

  @inproceedings{DBLP:conf/pkdd/StubbemannS23,
  author = {Stubbemann, Maximilian and Stumme, Gerd},
  booktitle = {European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML-PKDD 2023},
  keywords = {itegpub},
  pages = {177--192},
  publisher = {Springer},
  series = {Lecture Notes in Computer Science},
  title = {The Mont Blanc of Twitter: Identifying Hierarchies of Outstanding Peaks in Social Networks},
  volume = 14171,
  year = 2023
  }
  

  %0 Conference Paper
  %1 DBLP:conf/pkdd/StubbemannS23
  %A Stubbemann, Maximilian
  %A Stumme, Gerd
  %B European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML-PKDD 2023
  %D 2023
  %I Springer
  %P 177--192
  %R 10.1007/978-3-031-43418-1\_11
  %T The Mont Blanc of Twitter: Identifying Hierarchies of Outstanding Peaks in Social Networks
  %U https://doi.org/10.1007/978-3-031-43418-1\_11
  %V 14171
  

• 1.

  Stubbemann, M., Stumme, G.: LG4AV: Combining Language Models and Graph Neural Networks for Author Verification. In: Bouadi, T., Fromont, E., and H{ü}llermeier, E. (eds.) Advances in Intelligent Data Analysis XX. pp. 315–326. Springer International Publishing, Cham (2022).

  URLBibTeXEndNote

  The verification of document authorships is important in various settings. Researchers are for example judged and compared by the amount and impact of their publications and public figures are confronted by their posts on social media. Therefore, it is important that authorship information in frequently used data sets is correct. The question whether a given document is written by a given author is commonly referred to as authorship verification (AV). While AV is a widely investigated problem in general, only few works consider settings where the documents are short and written in a rather uniform style. This makes most approaches impractical for bibliometric data. Here, authorships of scientific publications have to be verified, often with just abstracts and titles available. To this point, we present LG4AV which combines language models and graph neural networks for authorship verification. By directly feeding the available texts in a pre-trained transformer architecture, our model does not need any hand-crafted stylometric features that are not meaningful in scenarios where the writing style is, at least to some extent, standardized. By the incorporation of a graph neural network structure, our model can benefit from relations between authors that are meaningful with respect to the verification process.

  @inproceedings{10.1007/978-3-031-01333-1_25,
  abstract = {The verification of document authorships is important in various settings. Researchers are for example judged and compared by the amount and impact of their publications and public figures are confronted by their posts on social media. Therefore, it is important that authorship information in frequently used data sets is correct. The question whether a given document is written by a given author is commonly referred to as authorship verification (AV). While AV is a widely investigated problem in general, only few works consider settings where the documents are short and written in a rather uniform style. This makes most approaches impractical for bibliometric data. Here, authorships of scientific publications have to be verified, often with just abstracts and titles available. To this point, we present LG4AV which combines language models and graph neural networks for authorship verification. By directly feeding the available texts in a pre-trained transformer architecture, our model does not need any hand-crafted stylometric features that are not meaningful in scenarios where the writing style is, at least to some extent, standardized. By the incorporation of a graph neural network structure, our model can benefit from relations between authors that are meaningful with respect to the verification process.},
  address = {Cham},
  author = {Stubbemann, Maximilian and Stumme, Gerd},
  booktitle = {Advances in Intelligent Data Analysis XX},
  editor = {Bouadi, Tassadit and Fromont, Elisa and H{ü}llermeier, Eyke},
  keywords = {itegpub},
  pages = {315--326},
  publisher = {Springer International Publishing},
  title = {LG4AV: Combining Language Models and Graph Neural Networks for Author Verification},
  year = 2022
  }
  

  %0 Conference Paper
  %1 10.1007/978-3-031-01333-1_25
  %A Stubbemann, Maximilian
  %A Stumme, Gerd
  %B Advances in Intelligent Data Analysis XX
  %C Cham
  %D 2022
  %E Bouadi, Tassadit
  %E Fromont, Elisa
  %E H{ü}llermeier, Eyke
  %I Springer International Publishing
  %P 315--326
  %T LG4AV: Combining Language Models and Graph Neural Networks for Author Verification
  %U https://link.springer.com/chapter/10.1007/978-3-031-01333-1_25
  %X The verification of document authorships is important in various settings. Researchers are for example judged and compared by the amount and impact of their publications and public figures are confronted by their posts on social media. Therefore, it is important that authorship information in frequently used data sets is correct. The question whether a given document is written by a given author is commonly referred to as authorship verification (AV). While AV is a widely investigated problem in general, only few works consider settings where the documents are short and written in a rather uniform style. This makes most approaches impractical for bibliometric data. Here, authorships of scientific publications have to be verified, often with just abstracts and titles available. To this point, we present LG4AV which combines language models and graph neural networks for authorship verification. By directly feeding the available texts in a pre-trained transformer architecture, our model does not need any hand-crafted stylometric features that are not meaningful in scenarios where the writing style is, at least to some extent, standardized. By the incorporation of a graph neural network structure, our model can benefit from relations between authors that are meaningful with respect to the verification process.
  %@ 978-3-031-01333-1
  

• 1.

  Koopmann, T., Stubbemann, M., Kapa, M., Paris, M., Buenstorf, G., Hanika, T., Hotho, A., Jäschke, R., Stumme, G.: Proximity dimensions and the emergence of collaboration: a HypTrails study on German AI research. Scientometrics. (2021). https://doi.org/10.1007/s11192-021-03922-1.

  URLBibTeXEndNote

  Creation and exchange of knowledge depends on collaboration. Recent work has suggested that the emergence of collaboration frequently relies on geographic proximity. However, being co-located tends to be associated with other dimensions of proximity, such as social ties or a shared organizational environment. To account for such factors, multiple dimensions of proximity have been proposed, including cognitive, institutional, organizational, social and geographical proximity. Since they strongly interrelate, disentangling these dimensions and their respective impact on collaboration is challenging. To address this issue, we propose various methods for measuring different dimensions of proximity. We then present an approach to compare and rank them with respect to the extent to which they indicate co-publications and co-inventions. We adapt the HypTrails approach, which was originally developed to explain human navigation, to co-author and co-inventor graphs. We evaluate this approach on a subset of the German research community, specifically academic authors and inventors active in research on artificial intelligence (AI). We find that social proximity and cognitive proximity are more important for the emergence of collaboration than geographic proximity.

  @article{koopmann2021proximity,
  abstract = {Creation and exchange of knowledge depends on collaboration. Recent work has suggested that the emergence of collaboration frequently relies on geographic proximity. However, being co-located tends to be associated with other dimensions of proximity, such as social ties or a shared organizational environment. To account for such factors, multiple dimensions of proximity have been proposed, including cognitive, institutional, organizational, social and geographical proximity. Since they strongly interrelate, disentangling these dimensions and their respective impact on collaboration is challenging. To address this issue, we propose various methods for measuring different dimensions of proximity. We then present an approach to compare and rank them with respect to the extent to which they indicate co-publications and co-inventions. We adapt the HypTrails approach, which was originally developed to explain human navigation, to co-author and co-inventor graphs. We evaluate this approach on a subset of the German research community, specifically academic authors and inventors active in research on artificial intelligence (AI). We find that social proximity and cognitive proximity are more important for the emergence of collaboration than geographic proximity.},
  author = {Koopmann, Tobias and Stubbemann, Maximilian and Kapa, Matthias and Paris, Michael and Buenstorf, Guido and Hanika, Tom and Hotho, Andreas and Jäschke, Robert and Stumme, Gerd},
  journal = {Scientometrics},
  keywords = {regio},
  title = {Proximity dimensions and the emergence of collaboration: a HypTrails study on German AI research},
  year = 2021
  }
  

  %0 Journal Article
  %1 koopmann2021proximity
  %A Koopmann, Tobias
  %A Stubbemann, Maximilian
  %A Kapa, Matthias
  %A Paris, Michael
  %A Buenstorf, Guido
  %A Hanika, Tom
  %A Hotho, Andreas
  %A Jäschke, Robert
  %A Stumme, Gerd
  %D 2021
  %J Scientometrics
  %R 10.1007/s11192-021-03922-1
  %T Proximity dimensions and the emergence of collaboration: a HypTrails study on German AI research
  %U https://doi.org/10.1007/s11192-021-03922-1
  %X Creation and exchange of knowledge depends on collaboration. Recent work has suggested that the emergence of collaboration frequently relies on geographic proximity. However, being co-located tends to be associated with other dimensions of proximity, such as social ties or a shared organizational environment. To account for such factors, multiple dimensions of proximity have been proposed, including cognitive, institutional, organizational, social and geographical proximity. Since they strongly interrelate, disentangling these dimensions and their respective impact on collaboration is challenging. To address this issue, we propose various methods for measuring different dimensions of proximity. We then present an approach to compare and rank them with respect to the extent to which they indicate co-publications and co-inventions. We adapt the HypTrails approach, which was originally developed to explain human navigation, to co-author and co-inventor graphs. We evaluate this approach on a subset of the German research community, specifically academic authors and inventors active in research on artificial intelligence (AI). We find that social proximity and cognitive proximity are more important for the emergence of collaboration than geographic proximity.
  

• 1.

  Stubbemann, M., Hanika, T., Stumme, G.: Orometric Methods in Bounded Metric Data. In: Berthold, M.R., Feelders, A., and Krempl, G. (eds.) Advances in Intelligent Data Analysis {XVIII} - 18th International Symposium on Intelligent Data Analysis, {IDA} 2020, Konstanz, Germany, April 27-29, 2020, Proceedings. pp. 496–508. Springer (2020). https://doi.org/10.1007/978-3-030-44584-3\_39.

  URLBibTeXEndNote

  @inproceedings{DBLP:conf/ida/StubbemannHS20,
  author = {Stubbemann, Maximilian and Hanika, Tom and Stumme, Gerd},
  booktitle = {Advances in Intelligent Data Analysis {XVIII} - 18th International Symposium on Intelligent Data Analysis, {IDA} 2020, Konstanz, Germany, April 27-29, 2020, Proceedings},
  editor = {Berthold, Michael R. and Feelders, Ad and Krempl, Georg},
  keywords = {itegpub},
  pages = {496--508},
  publisher = {Springer},
  series = {Lecture Notes in Computer Science},
  title = {Orometric Methods in Bounded Metric Data},
  volume = 12080,
  year = 2020
  }
  

  %0 Conference Paper
  %1 DBLP:conf/ida/StubbemannHS20
  %A Stubbemann, Maximilian
  %A Hanika, Tom
  %A Stumme, Gerd
  %B Advances in Intelligent Data Analysis {XVIII} - 18th International Symposium on Intelligent Data Analysis, {IDA} 2020, Konstanz, Germany, April 27-29, 2020, Proceedings
  %D 2020
  %E Berthold, Michael R.
  %E Feelders, Ad
  %E Krempl, Georg
  %I Springer
  %P 496--508
  %R 10.1007/978-3-030-44584-3\_39
  %T Orometric Methods in Bounded Metric Data
  %U https://doi.org/10.1007/978-3-030-44584-3\_39
  %V 12080
  

• 1.

  Dürrschnabel, D., Hanika, T., Stubbemann, M.: FCA2VEC: Embedding Techniques for Formal Concept Analysis, http://arxiv.org/abs/1911.11496, (2019).

  URLBibTeXEndNote

  Embedding large and high dimensional data into low dimensional vector spaces is a necessary task to computationally cope with contemporary data sets. Superseding latent semantic analysis recent approaches like word2vec or node2vec are well established tools in this realm. In the present paper we add to this line of research by introducing fca2vec, a family of embedding techniques for formal concept analysis (FCA). Our investigation contributes to two distinct lines of research. First, we enable the application of FCA notions to large data sets. In particular, we demonstrate how the cover relation of a concept lattice can be retrieved from a computational feasible embedding. Secondly, we show an enhancement for the classical node2vec approach in low dimension. For both directions the overall constraint of FCA of explainable results is preserved. We evaluate our novel procedures by computing fca2vec on different data sets like, wiki44 (a dense part of the Wikidata knowledge graph), the Mushroom data set and a publication network derived from the FCA community.

  @misc{durrschnabel2019fca2vec,
  abstract = {Embedding large and high dimensional data into low dimensional vector spaces is a necessary task to computationally cope with contemporary data sets. Superseding latent semantic analysis recent approaches like word2vec or node2vec are well established tools in this realm. In the present paper we add to this line of research by introducing fca2vec, a family of embedding techniques for formal concept analysis (FCA). Our investigation contributes to two distinct lines of research. First, we enable the application of FCA notions to large data sets. In particular, we demonstrate how the cover relation of a concept lattice can be retrieved from a computational feasible embedding. Secondly, we show an enhancement for the classical node2vec approach in low dimension. For both directions the overall constraint of FCA of explainable results is preserved. We evaluate our novel procedures by computing fca2vec on different data sets like, wiki44 (a dense part of the Wikidata knowledge graph), the Mushroom data set and a publication network derived from the FCA community.},
  author = {Dürrschnabel, Dominik and Hanika, Tom and Stubbemann, Maximilian},
  keywords = {itegpub},
  note = {cite arxiv:1911.11496Comment: 25 pages},
  title = {FCA2VEC: Embedding Techniques for Formal Concept Analysis},
  year = 2019
  }
  

  %0 Generic
  %1 durrschnabel2019fca2vec
  %A Dürrschnabel, Dominik
  %A Hanika, Tom
  %A Stubbemann, Maximilian
  %D 2019
  %T FCA2VEC: Embedding Techniques for Formal Concept Analysis
  %U http://arxiv.org/abs/1911.11496
  %X Embedding large and high dimensional data into low dimensional vector spaces is a necessary task to computationally cope with contemporary data sets. Superseding latent semantic analysis recent approaches like word2vec or node2vec are well established tools in this realm. In the present paper we add to this line of research by introducing fca2vec, a family of embedding techniques for formal concept analysis (FCA). Our investigation contributes to two distinct lines of research. First, we enable the application of FCA notions to large data sets. In particular, we demonstrate how the cover relation of a concept lattice can be retrieved from a computational feasible embedding. Secondly, we show an enhancement for the classical node2vec approach in low dimension. For both directions the overall constraint of FCA of explainable results is preserved. We evaluate our novel procedures by computing fca2vec on different data sets like, wiki44 (a dense part of the Wikidata knowledge graph), the Mushroom data set and a publication network derived from the FCA community.
  

Talks

• September 2023: The Mont Blanc of Twitter: Identifying Hierarchies of Outstanding Peaks in Social Networks, ECML/PKDD 2023
• September 2023: Intrinsic Dimensionality and Graph Learning, ISI Foundation Turin
• January 2023: Topological Data Analysis and Neural Networks, Dagstuhl Workshop on Concept Lattice Based Topological Data Analysis and Reasoning
• April 2022: LG4AV: Combining Language Models and Graph Neural Networks for Author Verification, Symposium on Intelligent Data Analysis 2022
• July 2021: Dimensionen von Nähe und ihr Einfluss auf die Entstehung von Kollaboration, ITeG Brown Bag Seminar
• April 2020: Orometric Methods in Bounded Metric Data, Symposium on Intelligent Data Analysis 2020

Research Communication

• Podcast of the University of Kassel: How does ChatGpt work? (In German)
• Participation in a radio report on ChatGPT on the Hessian radio station HR4, broadcast on 23.03.2023 at 11:15 a.m.
• Inteview for the local magazine “Mein Kassel” (in German) :
  Current state and future of language models and their implications four our lives.
• External Lecturer at the Departement of Empirical Research on Schools and Teaching:
  ChatGPT et al. : Functionality, History and Consequences for Teaching (in German)
  

Reviewing

• Subreviewer: European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, September 19-23 2022
• Subreviewer: 20th International Semantic Web Conference, October 24-28 2021, Virtual Conference
• Subreviewer: 18th Extended Semantic Web Conference, June 6-10 2021, Hersonissos, Greece
• Subreviewer: 19th International Semantic Web Conference, November 1-6 2020, Virtual Conference
• Subreviewer: 25th International Conference on Conceptual Structures, September 18-21 2020, Bolzano, Italy
• Subreviewer: 24th European Conference on Artificial Intelligence, June 8-12 2020, Santiago de Compostela, Spain
• Subreviewer: 25th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, August 4 – 8, 2019, Anchorage, Alaska – USA
• Subreviewer: 15th International Conference on Formal Concept Analysis, June 25-28 2019 – Frankfurt, Germany 

Projects

REGIO (2018-2021)

REGIO was a joint project between the University of Kassel, the L3S Research Center Hannover, the HU Berlin and the University of Würzburg. Together, we established new methodologies and data sources which helped to better understand the impact of geographic and thematic proximity on the genesis and the success of interaction in science and R&D. Our findings indicate, that social and thematic proximity are the key factors for cooperation.

Dimension Curse Detector (2022-)

This project is led by Dr. Tom Hanika within the Loewe Exploration program. In this project, we investigate to which extent machine learning is influenced by the intrinsic dimensionality of data. For this, we quantify the Curse of Dimensionality which is strongly connected to the phenomenon of measure concentration. Within the project, we develop methods that allow to efficiently compute the intrinsic dimension of modern large-scale datasets.

Teaching

Together with Johannes Hirth, I developed a Clojure programming curse which first took place in summer 2019. The course is hold until today on a regular basis.

Wintersemester 2022

• Knowledge Discovery (Praktikum)

Sommersemester 2021

• Datenbanken

Wintersemester 2021

• Knowledge Discovery (Praktikum)

Sommersemester 2020

• Datenbanken
• Seminar: Ausgewählte Themen der Wissensverarbeitung

Winter 2019/20

• Soziale Netzwerkanalyse
• Labor Netzwerke
• Clojure-Kurs

Summer 2019

• Datenbanken
• Clojure-Kurs

Winter 2018/2019

• Knowledge Discovery (Praktikum)