Riemannian Optimization for Hyperbolic Prototypical Networks
Keywords: Hyperbolic Representation Learning, Prototype Learning, Image Classification
TL;DR: We propose a prototypical framework in hyperbolic spaces to improve the performances of the state-of-the-art in Prototype Learning and and to advance the research in hyperbolic representation learning.
Abstract: This paper addresses the utilization of hyperbolic geometry within a Prototype Learning framework. Specifically, we introduce Riemannian optimization for Hyperbolic Prototypical Networks (RHPN), a novel approach that leverages Prototype Learning on Riemannian manifolds applied to the Poincare' ball. RHPN capitalizes on the efficiency and effectiveness of updating prototypes during training, coupled with a regularization term crucial to boost the performances. We set up an extensive experimentation that shows that RHPN is able to outperform the state-of-the-art in Prototype Learning, both in low and high dimensions, extending the impact of hyperbolic spaces to a wider range of scenarios.
Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 10942
Loading