DiagViB-6: A Diagnostic Benchmark Suite for Vision Models in the Presence of Shortcut and Generalization Opportunities

Eulig, Elias; Saranrittichai, Piyapat; Mummadi, Chaithanya Kumar; Rambach, Kilian; Beluch, William; Shi, Xiahan; Fischer, Volker

doi:10.1109/iccv48922.2021.01048

Cited by 6 publications

(1 citation statement)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Similar learning setups have appeared in prior papers: "Cross-bias generalisation" (Bahng et al, 2020), "What if multiple features are predictive?" (Hermann & Lampinen, 2020), and "Zero generalization opportunities" (Eulig et al, 2021). While we fully acknowledge the conceptual similarities, we stress that our work presents the first dedicated study into the cue selection problem and the underlying mechanisms.…”

Section: Data Framework: Wcst-mlmentioning

confidence: 83%

Which Shortcut Cues Will DNNs Choose? A Study from the Parameter-Space Perspective

Scimeca¹,

Oh²,

Chun³

et al. 2021

Preprint

View full text Add to dashboard Cite

Deep neural networks (DNNs) often rely on easy-to-learn discriminatory features, or cues, that are not necessarily essential to the problem at hand. For example, ducks in an image may be recognized based on their typical background scenery, such as lakes or streams. This phenomenon, also known as shortcut learning, is emerging as a key limitation of the current generation of machine learning models. In this work, we introduce a set of experiments to deepen our understanding of shortcut learning and its implications. We design a training setup with several shortcut cues, named WCST-ML, where each cue is equally conducive to the visual recognition problem at hand. Even under equal opportunities, we observe that (1) certain cues are preferred to others, (2) solutions biased to the easy-to-learn cues tend to converge to relatively flat minima on the loss surface, and (3) the solutions focusing on those preferred cues are far more abundant in the parameter space. We explain the abundance of certain cues via their Kolmogorov (descriptional) complexity: solutions corresponding to Kolmogorov-simple cues are abundant in the parameter space and are thus preferred by DNNs. Our studies are based on the synthetic dataset DSprites and the face dataset UTKFace. In our WCST-ML, we observe that the inborn bias of models leans toward simple cues, such as color and ethnicity. Our findings emphasize the importance of active human intervention to remove the inborn model biases that may cause negative societal impacts.

show abstract

Section: Data Framework: Wcst-mlmentioning

confidence: 83%