Skip to main content
QUICK REVIEW

[论文解读] Fast Objective and Duality Gap Convergence for Non-convex Strongly-concave Min-max Problems.

Zhishuai Guo, Zhuoning Yuan|arXiv (Cornell University)|Jun 12, 2020
Sparse and Compressive Sensing Techniques参考文献 45被引用 9
一句话总结

该论文提出了一种基于邻近时间片的随机算法,用于解决非凸强凹型极小-极大问题,通过引入一种新颖的李雅普诺夫函数和一种较弱的Polyak-Łojasiewicz(PL)条件,实现了在原始目标函数差距和对偶性差距上的快速收敛。该方法建立了最优的O(1/ε)迭代复杂度,并在CheXpert基准上实现了0.922的AUC。

ABSTRACT

This paper focuses on stochastic methods for solving smooth non-convex strongly-concave min-max problems, which have received increasing attention due to their potential applications in deep learning (e.g., deep AUC maximization). However, most of the existing algorithms are slow in practice, and their analysis revolves around the convergence to a nearly stationary point. We consider leveraging the Polyak-Łojasiewicz (PL) condition to design faster stochastic algorithms with stronger convergence guarantee. Although PL condition has been utilized for designing many stochastic minimization algorithms, their applications for non-convex min-max optimization remains rare. In this paper, we propose and analyze proximal epoch-based methods, and establish fast convergence in terms of both {\bf the primal objective gap and the duality gap}. Our analysis is interesting in threefold: (i) it is based on a novel Lyapunov function that consists of the primal objective gap and the duality gap of a regularized function; (ii) it only requires a weaker PL condition for establishing the primal objective convergence than that required for the duality gap convergence; (iii) it yields the optimal dependence on the accuracy level $\epsilon$, i.e., $O(1/\epsilon)$. We also make explicit the dependence on the problem parameters and explore regions of weak convexity parameter that lead to improved dependence on condition numbers. Experiments on deep AUC maximization demonstrate the effectiveness of our methods. Our method (MaxAUC) achieved an AUC of 0.922 on private testing set on {\bf CheXpert competition}.

研究动机与目标

  • 解决现有随机算法在深度学习应用中针对非凸强凹型极小-极大问题收敛缓慢的问题。
  • 在仅关注驻点之外,进一步提升收敛保证,同时针对原始目标函数差距和对偶性差距进行优化。
  • 利用弱于以往要求的PL条件,实现更快的收敛速率,以确保对偶性差距收敛。
  • 在温和假设下,实现对精度ε的最优迭代复杂度依赖,具体为O(1/ε)。
  • 探索弱凸性区域中参数依赖关系和条件数的改进。

提出的方法

  • 提出一种专为非凸强凹型极小-极大问题设计的邻近时间片随机算法。
  • 引入一种新颖的李雅普诺夫函数,结合正则化函数的原始目标函数差距和对偶性差距。
  • 采用较弱的PL条件以确保原始目标函数收敛,该条件与确保对偶性差距收敛所需的更强条件不同。
  • 利用该李雅普诺夫函数推导收敛速率,表明对精度ε具有最优的O(1/ε)依赖性。
  • 分析问题参数的显式依赖关系,并识别出可改善条件数依赖性的弱凸性区域。
  • 将该方法应用于深度AUC最大化,从而提出MaxAUC算法,适用于实际部署。

实验结果

研究问题

  • RQ1我们能否在非凸强凹型极小-极大问题中,同时实现原始目标函数差距和对偶性差距的更快收敛?
  • RQ2与对偶性差距收敛相比,是否较弱的PL条件已足够保证原始目标函数收敛?
  • RQ3在温和假设下,能否为这类问题实现最优的O(1/ε)迭代复杂度?
  • RQ4问题参数和条件数在弱凸性区域中如何影响收敛速率?
  • RQ5所提方法能否在真实深度学习任务(如AUC最大化)中优于现有方法?

主要发现

  • 所提算法在原始目标函数差距和对偶性差距收敛方面均实现了最优的O(1/ε)迭代复杂度。
  • 新颖的李雅普诺夫函数通过结合正则化函数的原始目标函数差距和对偶性差距,实现了更紧致的分析。
  • 在弱于对偶性差距收敛所需条件的PL条件下,原始目标函数收敛可被保证。
  • 该方法在特定弱凸性参数区域中,对条件数的依赖关系得到改善。
  • 在CheXpert竞赛中,MaxAUC方法在私有测试集上实现了0.922的AUC,展现出强劲的实证性能。

更好的研究,从现在开始

从阅读论文到最终审阅,大幅缩短您的研究时间。

无需绑定信用卡

本解读由 AI 生成,并经人工编辑审核。