Skip to Content

Werner Zellinger

Werner Zellinger's picture

PhD student since 2016
Master's Degree (Dipl.-Ing.) in Computer Mathematics at JKU Linz (2016)
Industrial Researcher at Software Competence Center Hagenberg since 2012 FLLL/JKU +43 7236 3343 430

Research interests:
I'm interested in the theoretical analysis and design of novel machine learning algorithms. I'm working on regularization-based algorithms and theoretical performance guarantees for transfer learning and generative image modeling. Applications can be found in industrial manufacturing, analytical chemistry and stereoscopic video analysis.
PhD supervisor:


  • Robust unsupervised domain adaptation for neural networks via moment alignment [link]
    Werner Zellinger, Bernhard Moser, Thomas Grubinger, Edwin Lughofer, Thomas Natschläger, and Susanne Saminger-Platz.
    Information Sciences, 483:174–191, 2019.
  • Domain-invariant partial least squares regression [link]
    Ramin Nikzad-Langerodi, Werner Zellinger, Edwin Lughofer, and Susanne Saminger-Platz.
    Analytical Chemistry, 90 (2018), 6693-6701.
  • Improving visual discomfort prediction for stereoscopic images via disparity-based contrast [link]
    Werner Zellinger, and Bernhard Moser.
    Journal of Imaging Science and Technology, 59.6 (2015), 60401-1.


  • Mixture Density Generative Adversarial Networks [link]
    Hamid Eghbal-zadeh, Werner Zellinger, Gerhard Widmer.
    The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2019.
  • Mixture Density Generative Adversarial Networks [arXiv]
    Hamid Eghbal-zadeh, Werner Zellinger, Gerhard Widmer.
    Neural Information Processing Systems (NeurIPS) workshop on Bayesian Deep Learning, 2018.
  • Moment distances for comparing high-entropy distributions with application in domain adaptation
    Werner Zellinger, Hamid Eghbal-zadeh, Bernhard Moser, Michael Zwick, Edwin Lughofer, Thomas Natschläger, and Susanne Saminger-Platz.
    International Conference on Computational Statistics (COMPSTAT), 2018.
  • Central moment discrepancy for domain invariant representation learning [arXiv] [poster] [code] [discussion]
    Werner Zellinger, Thomas Grubinger, Edwin Lughofer, Thomas Natschläger, and Susanne Saminger-Platz.
    International Conference on Learning Representations (ICLR), 2017.
  • Linear optimization approach for depth range adaption of stereoscopic videos [link] [presentation]
    Werner Zellinger, Bernhard Moser, Ayadi Chouikhi, Florian Seitner, Matej Nezveda, and Margrit Gelautz.
    Stereoscopic Displays and Applications (SD&A), 2016.
  • On the optimization of material usage in power transformer manufacturing
    Georgios Chasparis, Werner Zellinger, Verena Haunschmid, Markus Riedenbauer, and Reinhard Stumptner.
    International Conference on Intelligent Systems (IS), 2016.
  • Trifocal system for high-quality inter-camera mapping and virtual view synthesis
    Florian Seitner, Matej Nezveda, Margrit Gelautz, Georg Braun, Christian Kapeller, Werner Zellinger, and Bernhard Moser.
    International Conference on 3D Imaging (IC3D), 2015.
  • Parallelization of algorithms for linear discrete optimization using paraphrase
    Michael Rossbory, and Werner Zellinger.
    International Workshop on Database and Expert Systems Applications (DEXA), 2013.


  • Central moment discrepancy - domain adaptation via moment alignment
    Werner Zellinger, Thomas Grubinger, Edwin Lughofer, Thomas Natschläger, and Susanne Saminger-Platz.
    The Machine Learning Summer School (MLSS), Max Planck Institute, Tübingen, 2017.


  • Models and analysis of visual discomfort measures for stereoscopic images [download]
    Werner Zellinger.
    Master's Thesis (Johannes Kepler University Linz), 2015.


  • Wovon Maschinen Träumen [link]
    Werner Zellinger, and Robert Pollak.
    Kinderuni (Johannes Kepler University Linz), 2017.