arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2102.09468cs.LGmath.OCstat.ML
收藏

Minimax优化交替梯度下降-上升算法的近优局部收敛性

Near-optimal Local Convergence of Alternating Gradient Descent-Ascent for Minimax Optimization

University of Toronto(多伦多大学);Princeton University(普林斯顿大学);Northeastern University(东北大学)

Guodong Zhang, Yuanhao Wang, Laurent Lessard, Roger Grosse

摘要:

光滑极小极大博弈通常通过同时或交替的梯度更新来进行。虽然交替更新算法在实践中被广泛使用,但为了便于分析,现有的理论分析大多集中在并发算法上。本文研究了极大极小对策中的交替梯度下降-上升(Alt-GDA)算法,并证明了Alt-GDA算法在许多情况下优于其同步算法~(Sim-GDA)。我们证明了Alt-GDA实现了一个接近最优的局部收敛速度强凸-强凹(SCSC)问题,而Sim-GDA收敛速度要慢得多。据我们所知,这是任何设置的\emph{first}结果,表明Alt-GDA比Sim-GDA收敛快一个常数以上。我们进一步适应积分二次约束(IQC)的理论,并表明Alt-GDA达到相同的速度\emph{globally}的一个子类的SCSC极大极小问题。从经验上讲,我们证明了交替更新可以显着加快GAN训练,并且乐观的使用只对同步算法有帮助。

英文摘要:

Smooth minimax games often proceed by simultaneous or alternating gradient updates. Although algorithms with alternating updates are commonly used in practice, the majority of existing theoretical analyses focus on simultaneous algorithms for convenience of analysis. In this paper, we study alternating gradient descent-ascent (Alt-GDA) in minimax games and show that Alt-GDA is superior to its simultaneous counterpart~(Sim-GDA) in many settings. We prove that Alt-GDA achieves a near-optimal local convergence rate for strongly convex-strongly concave (SCSC) problems while Sim-GDA converges at a much slower rate. To our knowledge, this is the \emph{first} result of any setting showing that Alt-GDA converges faster than Sim-GDA by more than a constant. We further adapt the theory of integral quadratic constraints (IQC) and show that Alt-GDA attains the same rate \emph{globally} for a subclass of SCSC minimax problems. Empirically, we demonstrate that alternating updates speed up GAN training significantly and the use of optimism only helps for simultaneous algorithms.

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。

↑