Steffen Illium, M.Sc.

Steffen Illium, M.Sc.

Lehrstuhl für Mobile und Verteilte Systeme

Ludwig-Maximilians-Universität München, Institut für Informatik

Oettingenstraße 67
80538 München

Zoom Personal Meeting Room:
http://zoom.us/my/steffen.illium

Discord:
Steffen Illium#7765

Raum E 109 - Derzeit nicht besetzt.

Telefon: +49 89 / 2180-9174 - Derzeit nicht besetzt.

Fax:

Mail: steffen.illium@ifi.lmu.de

Lehre

Hauptseminare TIMS / VTIMS
Praktikum Mobile und Verteilte Systeme (MSP)
Praktikum IOS
Python Crashkurs

Research topics

  • Active:
    Anomaly Detection; Machine Learning; Self-Replicating Neural Networks; Deep Audio Analysis; Memory Horizon in RNN
  • Past:
    Clustering; Geoinformatics; Trajectory Analysis

Theses currently in progress:

  • Lernen geeigneter Repräsentationen zur Anomalie-Erkennung in Audio-Daten durch unüberwachte Pseudo-Klassifikation
  • Hierarchical Transformer-based Encoder for Contextual Sentence Embeddings

Open Theses (MSc/BSc) / Einzel Praktika (EP)

  • Allgemeine Informationen:
    http://www.mobile.ifi.lmu.de/lehre/ausgeschriebene-abschlussarbeiten/
  • Open Thesis Topics:
    Acoustic Anomaly Detection
    Evaluation of Data Augmentation for Audio-Data
    Per Frequency / Frequency Bin Modules for Anomaly Detection & Classification
    Combining Wavenet with Anomaly Detection
    Approximating the Wave Function Collapse-Algorithm with a Growing Neural Cellular Automata
    PointClouds: Evaluation of Neural Network performance in Message Passing
  • Old:
    Defining a Metric for Movement Analysis (EP)
    Capturing Isovist in Constraint Free Space (EP)
    Alternative Route Generator for Indoor Environments (MSc / EP)
    Interpolation in agents weight space for varying tasks (BSc / MSc / EP)

Abgeschlossene Abschlussarbeiten

  • Additional tasks for self-replicating neural networks
  • Reachability analysis for a hypothetical car sharing service
  • Segmentierung und Klassifizierung von geometrischen Primitiven in 3D-Punktwolken mit neuronalen Netzen
  • Vergleichende Analyse generativer neuronaler Netze auf Datenbasis des Wave Function Collapse-Algorithmus
  • Lokalisierung von Leckstellen in Wasserverteilungsanlagen mit Deep Learning via Data Augmentation
  • Analyse unüberwachten Clusterings von Isovisten mittels Attention-Mechanismus
  • Audiosignalverarbeitung und Clustering von Audiosignalen in Wasserleitungen
  • Investigating Memory Horizons in Neural Memory Networks

Publikationen

2020

  • S. Illium, R. Müller, A. Sedlmeier, and C. Linnhoff-Popien, "Surgical Mask Detection with Convolutional Neural Networks and Data Augmentations on Spectrograms," in Proc. Interspeech 2020, 2020, pp. 2052-2056. doi:10.21437/Interspeech.2020-1692
    [BibTeX] [Download PDF]
    @inproceedings{Illium2020,
    author={Steffen Illium and Robert Müller and Andreas Sedlmeier and Claudia Linnhoff-Popien},
    title={{Surgical Mask Detection with Convolutional Neural Networks and Data Augmentations on Spectrograms}},
    year=2020,
    booktitle={Proc. Interspeech 2020},
    pages={2052--2056},
    doi={10.21437/Interspeech.2020-1692},
    url={http://dx.doi.org/10.21437/Interspeech.2020-1692}
    }
  • A. Sedlmeier, R. Müller, S. Illium, and C. Linnhoff-Popien, "Policy Entropy for Out-of-Distribution Classification," in Artificial Neural Networks and Machine Learning -- ICANN 2020, Cham, 2020, pp. 420-431.
    [BibTeX] [Abstract]
    One critical prerequisite for the deployment of reinforcement learning systems in the real world is the ability to reliably detect situations on which the agent was not trained. Such situations could lead to potential safety risks when wrong predictions lead to the execution of harmful actions. In this work, we propose PEOC, a new policy entropy based out-of-distribution classifier that reliably detects unencountered states in deep reinforcement learning. It is based on using the entropy of an agent's policy as the classification score of a one-class classifier. We evaluate our approach using a procedural environment generator. Results show that PEOC is highly competitive against state-of-the-art one-class classification algorithms on the evaluated environments. Furthermore, we present a structured process for benchmarking out-of-distribution classification in reinforcement learning.
    @InProceedings{sedlmeier2020peoc,
    author="Sedlmeier, Andreas
    and M{\"u}ller, Robert
    and Illium, Steffen
    and Linnhoff-Popien, Claudia",
    editor="Farka{\v{s}}, Igor
    and Masulli, Paolo
    and Wermter, Stefan",
    title="Policy Entropy for Out-of-Distribution Classification",
    booktitle="Artificial Neural Networks and Machine Learning -- ICANN 2020",
    year="2020",
    publisher="Springer International Publishing",
    address="Cham",
    pages="420--431",
    abstract="One critical prerequisite for the deployment of reinforcement learning systems in the real world is the ability to reliably detect situations on which the agent was not trained. Such situations could lead to potential safety risks when wrong predictions lead to the execution of harmful actions. In this work, we propose PEOC, a new policy entropy based out-of-distribution classifier that reliably detects unencountered states in deep reinforcement learning. It is based on using the entropy of an agent's policy as the classification score of a one-class classifier. We evaluate our approach using a procedural environment generator. Results show that PEOC is highly competitive against state-of-the-art one-class classification algorithms on the evaluated environments. Furthermore, we present a structured process for benchmarking out-of-distribution classification in reinforcement learning.",
    isbn="978-3-030-61616-8"
    }
  • S. Illium, P. A. Friese, R. Müller, and S. Feld, "What to do in the Meantime: A Service Coverage Analysis for Parked Autonomous Vehicles," AGILE: GIScience Series, vol. 1, p. 7, 2020. doi:10.5194/agile-giss-1-7-2020
    [BibTeX] [Download PDF]
    @Article{agile-giss-1-7-2020,
    AUTHOR = {Illium, S. and Friese, P. A. and M\"uller, R. and Feld, S.},
    TITLE = {What to do in the Meantime: A Service Coverage Analysis for Parked Autonomous Vehicles},
    JOURNAL = {AGILE: GIScience Series},
    VOLUME = {1},
    YEAR = {2020},
    PAGES = {7},
    URL = {https://agile-giss.copernicus.org/articles/1/7/2020/},
    DOI = {10.5194/agile-giss-1-7-2020}
    }
  • M. Friedrich, S. Illium, P. Fayolle, and C. Linnhoff-Popien, "A Hybrid Approach for Segmenting and Fitting Solid Primitives to 3D Point Clouds," in Proceedings of the 15th International Conference on Computer Graphics Theory and Applications (GRAPP), 2020.
    [BibTeX]
    @inproceedings{friedrich2020hybrid,
    title={A Hybrid Approach for Segmenting and Fitting Solid Primitives to 3D Point Clouds},
    author={Friedrich, Markus and Illium, Steffen and Fayolle, Pierre-Alain and Linnhoff-Popien, Claudia},
    booktitle={Proceedings of the 15th International Conference on Computer Graphics Theory and Applications (GRAPP)},
    volume={1},
    year={2020},
    organization={INSTICC},
    publisher={SCITEPRESS},
    }

2019

  • D. Elsner, S. Langer, F. Ritz, R. Mueller, and S. Illium, "Deep Neural Baselines for Computational Paralinguistics," Proc. Interspeech 2019, pp. 2388-2392, 2019.
    [BibTeX]
    @article{elsner2019deep,
    title={Deep Neural Baselines for Computational Paralinguistics},
    author={Elsner, Daniel and Langer, Stefan and Ritz, Fabian and Mueller, Robert and Illium, Steffen},
    journal={Proc. Interspeech 2019},
    pages={2388--2392},
    year={2019}
    }
  • R. Müller, S. Langer, F. Ritz, C. Roch, S. Illium, and C. Linnhoff-Popien, "Soccer Team Vectors," in Joint European Conference on Machine Learning and Knowledge Discovery in Databases, 2019, pp. 247-257.
    [BibTeX]
    @inproceedings{mueller2019steve,
    title={Soccer Team Vectors},
    author={M{\"u}ller, Robert and Langer, Stefan and Ritz, Fabian and Roch, Christoph and Illium, Steffen and Linnhoff-Popien, Claudia},
    booktitle={Joint European Conference on Machine Learning and Knowledge Discovery in Databases},
    pages={247--257},
    year={2019},
    organization={Springer}
    }
  • T. Gabor, S. Illium, A. Mattausch, L. Belzner, and C. Linnhoff-Popien, "Self-Replication in Neural Networks," in Conference on Artificial Life (ALIFE 2019), 2019.
    [BibTeX]
    @InProceedings{gabor2019self-replication,
    author = {Thomas Gabor and Steffen Illium and Andy Mattausch and Lenz Belzner and Claudia Linnhoff-Popien},
    title = {Self-Replication in Neural Networks},
    booktitle = {Conference on Artificial Life (ALIFE 2019)},
    year = {2019},
    owner = {tgabor},
    }

2018

  • S. Illium, T. Gabor, and T. Phan, "Bayesian Variational Optimization in Sensor Networks," in 17. GI/ITG KuVS Fachgespräch Sensornetze 13. & 14. September 2018, Braunschweig: Technical Report, Braunschweig, 2018. doi:10.24355/dbbs.084-201809121401-1
    [BibTeX] [Download PDF]
    @InProceedings{illium2018bayesian,
    author = {Illium, Steffen and Gabor, Thomas and Phan, Thomy},
    title = {Bayesian Variational Optimization in Sensor Networks},
    booktitle = {17. GI/ITG KuVS Fachgespr{\"a}ch Sensornetze 13. \& 14. September 2018, Braunschweig: Technical Report},
    year = {2018},
    editor = {Wolf , Lars and B{\"u}sching, Felix},
    address = {Braunschweig},
    doi = {10.24355/dbbs.084-201809121401-1},
    owner = {sillium},
    url = {https://publikationsserver.tu-braunschweig.de/receive/dbbs_mods_00065986},
    }
  • S. Feld, S. Illium, A. Sedlmeier, and L. Belzner, "Trajectory annotation using sequences of spatial perception," in 26th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (SIGSPATIAL ’18), 2018, pp. 1-10. doi:https://doi.org/10.1145/3274895.3274968
    [BibTeX]
    @InProceedings{feld2018trajectory,
    author = {Feld, Sebastian and Illium, Steffen and Sedlmeier, Andreas and Belzner, Lenz},
    title = {Trajectory annotation using sequences of spatial perception},
    booktitle = {26th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (SIGSPATIAL ’18)},
    year = {2018},
    pages = {1--10},
    organization = {ACM},
    doi = {https://doi.org/10.1145/3274895.3274968},
    }