arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~

高校专区

Imperial College London(帝国理工学院)

2026-03-04 至 2026-03-04 共收录 5
2603.01073 2026-03-04 cs.CV

Flow Matching-enabled Test-Time Refinement for Unsupervised Cardiac MR Registration

基于流匹配的无监督心脏磁共振成像测试时细化

Yunguan Fu, Wenjia Bai, Wen Yan, Matthew J Clarkson, Rhodri Huw Davies, Yipeng Hu

机构 * University College London(伦敦大学学院) InstaDeep Imperial College London(伦敦帝国学院)

AI总结 FlowReg 通过流匹配框架实现高效的无监督心脏 MR 配准,无需预训练模型,提升配准精度并减少误差

详情

展开后加载摘要…

URL PDF HTML 收藏
2603.02452 2026-03-04 cs.LG cs.AI stat.ML

Manifold Aware Denoising Score Matching (MAD)

面向流形的去噪分数匹配(MAD)

Alona Levy-Jurgenson, Alvaro Prat, James Cuin, Yee Whye Teh

机构 * Department of Statistics University of Oxford(牛津大学统计系) Department of Mathematics, Imperial College London, London, United Kingdom(伦敦帝国理工学院数学系)

AI总结 本文提出了一种面向流形的去噪分数匹配方法,通过分解分数函数来隐式考虑流形,从而减少计算负担并提高效率。

详情

展开后加载摘要…

URL PDF HTML 收藏
2602.12274 2026-03-04 cs.LG physics.geo-ph

Function-Space Decoupled Diffusion for Forward and Inverse Modeling in Carbon Capture and Storage

函数空间解耦扩散用于碳捕集与封存的正反演建模

Xin Ju, Jiachen Yao, Anima Anandkumar, Sally M. Benson, Gege Wen

机构 * Stanford University(斯坦福大学) California Institute of Technology(加州理工学院) Imperial College London(伦敦帝国学院)

AI总结 函数空间解耦扩散方法在碳捕集与封存的正反演建模中实现高效且物理一致的参数恢复与数据同化。

Comments Accepted to ICLR AI&PDE Workshop

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.09143 2026-03-04 cs.LG cs.NA math.NA physics.comp-ph

Discrete Solution Operator Learning for Geometry-Dependent PDEs

几何依赖偏微分方程的离散解算子学习

Jinshuai Bai, Haolin Li, Zahra Sharif Khodaei, M. H. Aliabadi, YuanTong Gu, Xi-Qiao Feng

机构 * Institute of Biomechanics and Medical Engineering Applied Mechanics Laboratory (AML) Tsinghua University(生物力学与医学工程研究所应用力学实验室(AML)清华大学) Department of Aeronautics Imperial College London(航空航天系帝国理工学院伦敦) School of Mechanical, Medical, and Process Engineering Queensland University of Technology(机械、医学与工艺工程学院昆士兰理工大学)

AI总结 DiSOL通过学习离散求解过程来处理几何依赖的偏微分方程,实现稳定且准确的预测。

Comments 15 pages main text, 42 pages SI

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.00272 2026-03-04 cs.LG cs.AI cs.CR

WARP: Weight Teleportation for Attack-Resilient Unlearning Protocols

WARP:用于攻击鲁棒删除学习协议的权重传送

Mohammad M Maheri, Xavier Cadet, Peter Chin, Hamed Haddadi

机构 * Imperial College London(伦敦帝国学院) Dartmouth College(达特茅斯学院)

AI总结 WARP通过利用神经网络对称性减少遗忘集梯度能量和参数分散,提升删除学习协议的攻击鲁棒性,有效降低对抗优势。

Comments This work has been accepted for publication at the International Conference on Learning Representations (ICLR) 2026 (to appear)

详情

展开后加载摘要…

URL PDF HTML 收藏