Tag
This paper introduces a framework for E3-equivariant uncertainty quantification in tensor-valued geometric learning, ensuring symmetric positive-definite covariances via matrix exponentiation and proposing a robust Log-Euclidean Equivariant Scoring Objective.