Research Focus

My research focuses on the representation and quantification of uncertainty in machine learning — that is, on how a predictive model should encode what it does not know, and how that ignorance should be measured. On the representation side, I am particularly interested in ensembling methods and in credal-set representations, in which a prediction is given not by a single probability distribution but by a convex set of such distributions. On the quantification side, I work on uncertainty measures derived from proper scoring rules, which provide a principled basis for assessing predictive uncertainty and, in particular, for decomposing it into aleatoric and epistemic components.

Selected Publications

  • Paul Hofman, Timo Löhr, Maximilian Muschalik, Yusuf Sale, Eyke Hüllermeier
    Efficient Credal Prediction through Decalibration
    arXiv preprint arXiv:2603.08495
  • Paul Hofman, Yusuf Sale, Eyke Hüllermeier (2026)
    Uncertainty Quantification for Machine Learning: One Size Does Not Fit All(bib)(pdf)
    The 40th Annual AAAI Conference on Artificial Intelligence, 20. - 27. January 2026, Singapore
  • Timo Löhr, Paul Hofman, Felix Mohr, Eyke Hüllermeier
    Credal Prediction based on Relative Likelihood
    NeurIPS2025 (Spotlight)
  • Paul Hofman, Yusuf Sale, Eyke Hüllermeier
    Quantifying aleatoric and epistemic uncertainty: A credal approach
    ICML 2024 Workshop on Structured Probabilistic Inference & Generative Modeling