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~~
arXiv 2307.01166cs.LGmath.AP
收藏

通过耦合梯度流研究交互智能体的策略性分布偏移

Strategic Distribution Shift of Interacting Agents via Coupled Gradient Flows

California Institute of Technology(加州理工学院);University of Washington(华盛顿大学)

Lauren Conger, Franca Hoffmann, Eric Mazumdar, Lillian Ratliff

AI总结:

本文提出耦合偏微分方程框架刻画学习算法与分布之间的反馈动态,并在合作与竞争两类场景下证明梯度下降再训练渐近收敛到稳态,实验表明该方法能捕捉极化等复杂分布偏移现象。

摘要:

我们提出一个新的框架,用于分析现实系统中分布偏移的动态,该框架刻画了学习算法与其部署所依赖的分布之间的反馈回路。先前的工作大多将反馈引起的分布偏移建模为对抗性的,或采用过于简化的分布偏移结构。相比之下,我们提出一个耦合的偏微分方程模型,通过考虑由于对算法决策的策略性响应、非局部内生群体交互以及其他外生分布偏移来源所产生的复杂动态,来捕捉分布随时间的细粒度变化。我们考虑机器学习中两种常见场景:具有信息不对称的合作场景,以及学习者面对策略性用户的竞争场景。对于这两种场景,当算法通过梯度下降进行再训练时,我们证明了再训练过程在有限维和无限维下均渐近收敛到稳态,并给出了以模型参数表示的显式收敛速率。为此,我们推导了关于耦合偏微分方程收敛性的新结果,扩展了已知的多物种系统结论。在实验上,我们表明我们的方法能够捕捉到已有文献充分记录的分布偏移形式,如极化和差异性影响,而这些是更简单的模型无法捕捉的。

英文摘要:

We propose a novel framework for analyzing the dynamics of distribution shift in real-world systems that captures the feedback loop between learning algorithms and the distributions on which they are deployed. Prior work largely models feedback-induced distribution shift as adversarial or via an overly simplistic distribution-shift structure. In contrast, we propose a coupled partial differential equation model that captures fine-grained changes in the distribution over time by accounting for complex dynamics that arise due to strategic responses to algorithmic decision-making, non-local endogenous population interactions, and other exogenous sources of distribution shift. We consider two common settings in machine learning: cooperative settings with information asymmetries, and competitive settings where a learner faces strategic users. For both of these settings, when the algorithm retrains via gradient descent, we prove asymptotic convergence of the retraining procedure to a steady-state, both in finite and in infinite dimensions, obtaining explicit rates in terms of the model parameters. To do so we derive new results on the convergence of coupled PDEs that extends what is known on multi-species systems. Empirically, we show that our approach captures well-documented forms of distribution shifts like polarization and disparate impacts that simpler models cannot capture.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑