M Sc Lukas Huber

E-Mail
lukas.s.huber@unibe.ch
Büro
A 370
Postadresse
University of Bern
Department of Psychology
Fabrikstrasse 8
3012 Bern

Research interests

  • Visual object recognition
  • Perceptual categorisation
  • Mental Imagery
  • Computer vision
  • Application of multidimensional scaling methods

Google Scholar

Successful recognition and categorisation both demand the availability of robust mental representations of objects. In contrast to machines, humans learn such representations from relatively little visual input. In his PhD project, Lukas uses a combination of psychophysical experiments and comparison-based machine learning methods to investigate the process in which such representations are learned. More specifically, he studies how brain-internal generative processes, such as mental imagery, may act as a biological form of data augmentation that facilitates the data-efficient acquisition of robust object representations.
 
Lukas received his B.Sc. and M.Sc. in Psychology from the University of Bern, Switzerland. His PhD project, "Creating to perceive: The role of mental imagery in visual object recognition and categorisation", is founded by a career grant from the Swiss National Science Foundation (SNSF) and is jointly supervised by  Felix Wichmann (Neural Information Processing, University of Tübingen) and Fred Mast (Cognitive Psychology, Perception and Research Methods, University of Bern).

Publications

Upcoming talks:

"Using Conceptual Scaling to Map How Individuals Understand Abstract Concepts"
@ the TeaP Symposium: Toward a Unified View of Categorization: Linking Cognitive Mechanisms and Social Consequences, University of Tübingen—March 2026

Peer reviewed publications:

Huber, L. S., Künstle, D. E., & Reuter, K. (2026). Tracing truth through conceptual scaling. Cognition, 266, 106321. https://www.sciencedirect.com/science/article/pii/S0010027725002628

Huber, L. S., Mast, F. W., & Wichmann, F. A. Comparing supervised learning dynamics: Deep neural networks match human data efficiency but show a generalisation lag. In ICLR 2024 Workshop on Representational Alignment. https://openreview.net/pdf?id=yb9LLnUdqU

Lerch, L.*, Huber, L. S.*, Kamath, A., Pöllinger, A., Pahud de Mortanges, A., Obmann, V. C., ... & Reyes, M. (2024). DreamOn: a data augmentation strategy to narrow the robustness gap between expert radiologists and deep learning classifiers. Frontiers in Radiology, 4, 1420545. https://www.frontiersin.org/journals/radiology/articles/10.3389/fradi.2024.1420545/full

Deperrois, N., Petrovici, M. A., Jordan, J., Huber, L. S., & Senn, W. (2024). How adversarial REM dreams may facilitate creativity, and why we become aware of them. Clinical and Translational Neuroscience, 8(2), 21. https://www.mdpi.com/2514-183X/8/2/21

Huber, L. S., Geirhos, R., & Wichmann, F. A. (2023). The developmental trajectory of object recognition robustness: children are like small adults but unlike big deep neural networks. Journal of vision, 23(7), 4-4. https://jov.arvojournals.org/article.aspx?articleid=2791281

Huber, L. S., Reuter, K., & Cacchione, T. (2022). Children and adults don’t think they are free: A skeptical look at agent causationism. Advances in Experimental Philosophy of Causation, 189.  https://www.bloomsburycollections.com/book/advances-in-experimental-philosophy-of-causation/ch9-children-and-adults-don-t-think-they-are-free-a-skeptical-look-at-agent-causationism

Huber, L. S., Geirhos, R., & Wichmann, F. A. (2021). A four-year-old can outperform ResNet--50: Out-of-distribution robustness may not require large-scale experience. In SVRHM 2021 Workshop @ NeurIPS. https://openreview.net/pdf?id=7yMg2rS9N5I

Invited talks:

"Robustness gap between humans and artificial neural networks"
DiCarlo Lab, Massachusetts Institute of Technology, USA—May 2025

"Do we have to choose between behavioral and biological alignment? CNNs may be behaviorally more human-like than we thought"
Andreas Tolias Lab, Stanford University, USA—May 2025

"Learning robust representations of novel objects"
AFC Lab / CARLA Talks, University of Lausanne, Switzerland—September 2024

"Introducing Conceptual Scaling for Experimental Philosophy"
Experimental Philosophy Colloquium, University of Zürich, Switzerland—June 2023

Conference contributions:

Huber, L. S., Mast, F. W., & Wichmann, F. A. (2024; May). Contrasting learning dynamics: Immediate generalisation in humans and generalisation lag in deep neural networks. 
Poster @ Annual Meeting of Vision Sciences Scociety, Florida. 
https://jov.arvojournals.org/article.aspx?articleid=2802067

Huber, L. S., Mast, F. W., & Wichmann, F. A. (2024, May). Comparing supervised learning dynamics: Deep neural networks match human data efficiency but show a generalisation lag.
Poster @ ICLR 2024 Workshop on Representational Alignment, Vienna.

Huber, L. S., Künstle, D., & Reuter, K. (2024, June). Tracing Truth Through Conceptual Scaling: Mapping People’s Understanding of Abstract Concepts.
Talk @ European Experimental Philosophy Conference, Kraków

Huber, L. S., Geirhos, R., & Wichmann, F. A. (2021). The developmental trajectory of object recognition robustness: comparing children, adults, and CNNs.
Talk @ Annual Meeting of Vision Sciences Scociety, Florida. 
https://jov.arvojournals.org/article.aspx?articleid=2776987 
https://www.youtube.com/watch?v=CLnykEu_k9w&t=164s