发表机构
Graduate School of Information Science and Technology, The University of Tokyo; Institute of Science Tokyo; Communication Science Laboratories, NTT, Inc.; Center for Advanced Intelligence Project, RIKEN（东京大学信息科学与技术研究生院; 东京科学大学; NTT公司通信科学实验室; 理化学研究所先进智能项目中心）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对决策式预测下梯度方法的分布与损失函数假设局限，本文提出带收敛保证的自适应梯度方法及实用变体，实验显示其收敛更快且更稳定。
AI 中文摘要
我们研究决策式预测下的优化问题，即部署模型会影响未来数据分布。针对该场景，已有多种梯度方法被提出，但它们通常假设特定的数据分布或损失函数，限制了实际适用性。为克服这些局限，我们提出一种梯度优化方法，在更弱的假设下具备收敛性保证。该方法通过有限差分显式估计诱导分布偏移，支持更高维度的优化，适用于更广泛的损失函数和数据分布类别；我们还提出一种实用变体，减少所需样本量。数值实验表明，所提算法相比现有方法收敛更快、稳定性更强。
英文摘要
We study optimization under performative prediction, where deploying a model affects the future data distribution. For this setting, several gradient-based approaches have been proposed. However, they typically assume specific data distributions or loss functions, which limit their practical applicability. To overcome these limitations, we propose a gradient-based optimization method with convergence guarantees under substantially weaker assumptions. Our method explicitly estimates the induced distribution shift through finite differences. It enables higher-dimensional optimization across broader classes of loss functions and data distributions. We also propose a practical variant that reduces the number of samples required. Numerical experiments demonstrate that our proposed algorithms converge faster and more consistently than existing ones.
Comments31pages, 2 figures
