Saved in:
Bibliographic Details
Main Authors: Dosi, Muskan, Chiranjeev, Chiranjeev, Thakral, Kartik, Vatsa, Mayank, Singh, Richa
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.10576
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916791960207360
author Dosi, Muskan
Chiranjeev, Chiranjeev
Thakral, Kartik
Vatsa, Mayank
Singh, Richa
author_facet Dosi, Muskan
Chiranjeev, Chiranjeev
Thakral, Kartik
Vatsa, Mayank
Singh, Richa
contents Do contemporary diffusion models preserve the class geometry of hyperspherical data? Standard diffusion models rely on isotropic Gaussian noise in the forward process, inherently favoring Euclidean spaces. However, many real-world problems involve non-Euclidean distributions, such as hyperspherical manifolds, where class-specific patterns are governed by angular geometry within hypercones. When modeled in Euclidean space, these angular subtleties are lost, leading to suboptimal generative performance. To address this limitation, we introduce HyperSphereDiff to align hyperspherical structures with directional noise, preserving class geometry and effectively capturing angular uncertainty. We demonstrate both theoretically and empirically that this approach aligns the generative process with the intrinsic geometry of hyperspherical data, resulting in more accurate and geometry-aware generative models. We evaluate our framework on four object datasets and two face datasets, showing that incorporating angular uncertainty better preserves the underlying hyperspherical manifold. Resources are available at: {https://github.com/IAB-IITJ/Harmonizing-Geometry-and-Uncertainty-Diffusion-with-Hyperspheres/}
format Preprint
id arxiv_https___arxiv_org_abs_2506_10576
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Harmonizing Geometry and Uncertainty: Diffusion with Hyperspheres
Dosi, Muskan
Chiranjeev, Chiranjeev
Thakral, Kartik
Vatsa, Mayank
Singh, Richa
Computer Vision and Pattern Recognition
Do contemporary diffusion models preserve the class geometry of hyperspherical data? Standard diffusion models rely on isotropic Gaussian noise in the forward process, inherently favoring Euclidean spaces. However, many real-world problems involve non-Euclidean distributions, such as hyperspherical manifolds, where class-specific patterns are governed by angular geometry within hypercones. When modeled in Euclidean space, these angular subtleties are lost, leading to suboptimal generative performance. To address this limitation, we introduce HyperSphereDiff to align hyperspherical structures with directional noise, preserving class geometry and effectively capturing angular uncertainty. We demonstrate both theoretically and empirically that this approach aligns the generative process with the intrinsic geometry of hyperspherical data, resulting in more accurate and geometry-aware generative models. We evaluate our framework on four object datasets and two face datasets, showing that incorporating angular uncertainty better preserves the underlying hyperspherical manifold. Resources are available at: {https://github.com/IAB-IITJ/Harmonizing-Geometry-and-Uncertainty-Diffusion-with-Hyperspheres/}
title Harmonizing Geometry and Uncertainty: Diffusion with Hyperspheres
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2506.10576