Spherical Gaussian
Spherical Gaussians are probability distributions modeling data points clustered around a central point on a sphere, finding applications in diverse fields like computer vision, machine learning, and signal processing. Current research focuses on improving the efficiency and accuracy of algorithms using spherical Gaussians, including advancements in Gaussian mixture models, anisotropic kernel methods, and their integration within neural networks such as transformers and graph convolutional networks. These improvements enhance the representation of complex data structures, leading to better performance in tasks such as point cloud completion, 3D rendering, and robust statistical estimation. The resulting advancements have significant implications for various applications, including image processing, traffic state estimation, and molecular property prediction.
Papers
SGAT4PASS: Spherical Geometry-Aware Transformer for PAnoramic Semantic Segmentation
Xuewei Li, Tao Wu, Zhongang Qi, Gaoang Wang, Ying Shan, Xi Li
Learning Representations on the Unit Sphere: Investigating Angular Gaussian and von Mises-Fisher Distributions for Online Continual Learning
Nicolas Michel, Giovanni Chierchia, Romain Negrel, Jean-François Bercher