Von Mises

The von Mises distribution, and its high-dimensional generalization, the von Mises-Fisher distribution, are finding increasing use in modeling circular and directional data across diverse scientific fields. Current research focuses on applying these distributions within various machine learning frameworks, including contrastive learning, mixture models, and Bayesian regression, often to improve model interpretability, uncertainty quantification, and robustness. This work is impacting fields ranging from causal discovery and molecular conformation prediction to mitigating bias in face recognition and enhancing the performance of deep metric learning algorithms. The ability to effectively model directional data with these distributions is proving crucial for advancing several areas of scientific inquiry and technological application.