arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2202.09688math.OCcs.LG
收藏

一种方差缩减的随机加速原始对偶法

A Variance-Reduced Stochastic Accelerated Primal Dual Algorithm

Rutgers Business School(罗格斯商学院);Penn State University(宾夕法尼亚州立大学)

Bugra Can, Mert Gurbuzbalaban, Necdet Serhat Aybat

摘要:

在这项工作中,我们考虑强凸强凹(SCSC)鞍点(SP)问题$\min_{x\in\mathbb{R}^{d_x}}\max_{y\in\mathbb{R}^{d_y}}f(x,y)$,其中$f$是$L$-光滑的,$f(.,y)$是$μ$-强凸的每个$y$,和$f(x,.)$是$μ$-强凹的每个$x$。在机器学习中,在稳健的经验风险最小化(ERM)的背景下,例如$\textit{distributionally robust}$ ERM,经常出现这样的问题,其中使用小批量数据点来估计部分梯度。假设我们可以访问无偏随机一阶预言,我们认为Zhang等人[2021]最近针对SCSC SP问题引入的随机加速原始对偶(SAPD)算法是一种针对梯度噪声的鲁棒方法。特别地,当动量参数被设置为零时,SAPD恢复众所周知的随机梯度下降上升(SGDA)作为特殊情况,并且当动量参数被适当地调谐时,即,将$κ\triangleq L/μ$依赖性从SGDA的$κ^2$改善为$κ$。我们提出了有效的方差减少战略SAPD的基础上Richardson-Romberg外推,并表明我们的方法在实践中和理论上提高SAPD。

英文摘要:

In this work, we consider strongly convex strongly concave (SCSC) saddle point (SP) problems $\min_{x\in\mathbb{R}^{d_x}}\max_{y\in\mathbb{R}^{d_y}}f(x,y)$ where $f$ is $L$-smooth, $f(.,y)$ is $μ$-strongly convex for every $y$, and $f(x,.)$ is $μ$-strongly concave for every $x$. Such problems arise frequently in machine learning in the context of robust empirical risk minimization (ERM), e.g. $\textit{distributionally robust}$ ERM, where partial gradients are estimated using mini-batches of data points. Assuming we have access to an unbiased stochastic first-order oracle we consider the stochastic accelerated primal dual (SAPD) algorithm recently introduced in Zhang et al. [2021] for SCSC SP problems as a robust method against gradient noise. In particular, SAPD recovers the well-known stochastic gradient descent ascent (SGDA) as a special case when the momentum parameter is set to zero and can achieve an accelerated rate when the momentum parameter is properly tuned, i.e., improving the $κ\triangleq L/μ$ dependence from $κ^2$ for SGDA to $κ$. We propose efficient variance-reduction strategies for SAPD based on Richardson-Romberg extrapolation and show that our method improves upon SAPD both in practice and in theory.

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。

↑