#constrained optimization

7 stories · 0 in 7 days · 0 in 30 days

SafeDiSTL 1 multi-agent reinforcement learning 1 reinforcement learning 1 regularization 1 robustness 1

Stories, products and papers tagged "constrained optimization", newest first.

Oct 12

Papers Builders11:03

Chinese contentSafeDiSTL：多智能体强化学习安全框架

SafeDiSTL解决了多智能体系统中如何公平分配约束责任的问题，在三个基准任务上显著提升安全性能。

#SafeDiSTL #multi-agent reinforcement learning #constrained optimization

arXiv cs.AI@Hongrui Zheng et al. (3)Source

Sep 14

Papers BuildersSelected10:00

Unified regularization for robust RL

This research provides a theoretical framework for improving the robustness of reinforcement learning models, which is a key challenge in developing reliable AI systems.

#reinforcement learning #regularization #robustness #constrained optimization

arXiv cs.LG@Amine Andam et al. (3)Source

Jul 29

Papers09:49

Chinese contentDynamic Constraint Learning：用于未知约束优化的数据驱动框架

这篇论文提出DCL方法，在约束函数未知时用局部代理高效优化，解决了实际难题，值得关注。

#Dynamic Constraint Learning #constrained optimization #data-driven #trust region

arXiv cs.LG@Ezgi Oztekin et al. (3)Source

Jul 14

Papers09:15

Chinese content约束在线凸优化中OGD+Projection算法的累积约束违反下界

这篇理论论文首次给出了OGD+Projection算法在高维约束优化中违反约束的下界，搞在线凸优化的人可以看看。

#OGD+Projection #COCO #online convex optimization #constrained optimization

arXiv cs.LG@Haricharan Balasundaram et al. (3)Source

Jun 19

Papers09:42

Chinese contentMAMO：面向多目标约束优化的多智能体系统

这篇论文提出MAMO，用多智能体RL自动调权重，解决约束优化中手动调参难题。

#MAMO #multi-agent systems #reinforcement learning #constrained optimization

arXiv cs.LG@Federica FilippiniSource

Jun 9

PapersSelected12:40

Chinese contentDri-MED：应对偏好漂移与约束的线性上下文赌博机算法

在线推荐系统常面临用户偏好漂移和基线约束的挑战，Dri-MED 为这类问题提供了理论扎实且效果显著的解决方案，做推荐系统或在线学习的团队值得关注其算法设计。

#online learning #constrained optimization #recommendation systems

arXiv cs.LG@Udvas Das et al. (4)Source

May 21

PapersSelected10:22

Chinese content自收缩性改进约束在线凸优化：CCV 从 O(√T log T) 降至 O(log T)

约束在线优化是机器学习中的核心问题，这篇论文用简洁的投影算法大幅降低了累积约束违反，做在线学习或凸优化理论的研究者值得关注，其自收缩性技巧可能启发更多改进。

#online convex optimization #constrained optimization #regret analysis

arXiv cs.LG@Dhruv Sarkar, Abhishek SinhaSource