HypLL: The Hyperbolic Learning Library Article Swipe
YOU?
·
· 2023
· Open Access
·
· DOI: https://doi.org/10.1145/3581783.3613462
· OA: W4380551115
Deep learning in hyperbolic space is quickly gaining traction in the fields\nof machine learning, multimedia, and computer vision. Deep networks commonly\noperate in Euclidean space, implicitly assuming that data lies on regular\ngrids. Recent advances have shown that hyperbolic geometry provides a viable\nalternative foundation for deep learning, especially when data is hierarchical\nin nature and when working with few embedding dimensions. Currently however, no\naccessible open-source library exists to build hyperbolic network modules akin\nto well-known deep learning libraries. We present HypLL, the Hyperbolic\nLearning Library to bring the progress on hyperbolic deep learning together.\nHypLL is built on top of PyTorch, with an emphasis in its design for\nease-of-use, in order to attract a broad audience towards this new and\nopen-ended research direction. The code is available at:\nhttps://github.com/maxvanspengler/hyperbolic_learning_library.\n
