Reproduce “Quantification of Credal Uncertainty: A Distance-Based Approach” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xabier Gonzalez-Garcia, Siu Lun Chau, Julian Rodemann, Michele Caprio, Krikamol Muandet, Humberto Bustince, Sébastien Destercke, Eyke Hüllermeier and 1 more
The authors introduce distance-based uncertainty measures for multiclass credal predictions, sets of plausible class-probability distributions. Using IPMs, they define total uncertainty, set-valued aleatoric uncertainty, and epistemic uncertainty through distances to certainty vertices and between distributions. With total variation, epistemic uncertainty achieved mean AUC 89.8 versus entropy’s 84.5 on five KEEL datasets with more than ten classes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.