arXivDaily arXiv每日学术速递 周一至周五更新

高校专区

University of Southern California(南加州大学)

2025-11-26 至 2025-11-26 共收录 2
2511.20168 2025-11-26 cs.LG cs.AI

On the Limits of Momentum in Decentralized and Federated Optimization

动量在去中心化和联邦优化中的局限性

Riccardo Zaccone, Sai Praneeth Karimireddy, Carlo Masone

机构 * Department of Computer and Control Engineering, Polytechnic of Turin(计算机与控制工程系,都灵理工学院) Thomas Lord Department of Computer Science, USC Viterbi School of Engineering(托马斯·劳德计算机科学系,USC维特里商学院)

AI总结 本文研究了动量在去中心化和联邦优化中的收敛性限制,证明在循环客户端参与下,动量无法克服统计异质性,且步长递减策略无法保证收敛。

Comments Accepted at the 17th Workshop on Optimization for Machine Learning (OPT@NeurIPS2025)

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.18617 2025-11-26 cs.RO cs.CV

AutoFocus-IL: VLM-based Saliency Maps for Data-Efficient Visual Imitation Learning without Extra Human Annotations

AutoFocus-IL:基于视觉语言模型的数据高效视觉模仿学习中的显著性图

Litian Gong, Fatemeh Bahrani, Yutai Zhou, Amin Banayeeanzade, Jiachen Li, Erdem Bıyık

机构 * Department of Electrical and Computer Engineering, University of California, Riverside, USA(电气与计算机工程系,加州大学河滨分校) Thomas Lord Department of Computer Science, University of Southern California, USA(汤姆斯·劳德计算机科学系,南加州大学)

AI总结 AutoFocus-IL通过视觉语言模型自动生成显著性图,提升视觉模仿学习的数据效率和泛化能力,无需额外人类标注。

Comments 8 pages, 6 figures. Code and datasets available at http://autofocus-il.github.io/

详情

展开后加载摘要…

URL PDF HTML 收藏