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 2302.09376stat.MLcs.LG
收藏

为什么参数平均在SGD中是有益的?一个目标平滑的视角

Why is parameter averaging beneficial in SGD? An objective smoothing perspective

Agency for Science, Technology and Research(新加坡科技研究局);Kyushu Institute of Technology(九州工业大学);New York University(纽约大学);Flatiron Institute(弗莱堡研究所)

Atsushi Nitanda, Ryuhei Kikuchi, Shugo Maeda, Denny Wu

AI总结:

本文从目标平滑角度证明平均SGD能优化平滑目标以避免尖锐局部最小值,并通过实验验证适当步长的参数平均显著提升SGD性能。

摘要:

通常观察到随机梯度下降(SGD)及其变体隐式地选择具有良好泛化性能的解;这种隐式偏差通常以最小值的锐度来表征。Kleinberg等人(2018)将此偏差与SGD的平滑效应联系起来,该效应通过使用随机梯度噪声的卷积来消除尖锐的局部最小值。我们遵循这一研究路线,研究常用的平均SGD算法,Izmailov等人(2018)在经验上观察到该算法偏好平坦的最小值,因此实现了更好的泛化。我们证明,在某些问题设置中,平均SGD能够有效优化平滑目标,从而避免尖锐的局部最小值。在实验中,我们验证了我们的理论,并表明采用适当步长的参数平均确实显著提升了SGD的性能。

英文摘要:

It is often observed that stochastic gradient descent (SGD) and its variants implicitly select a solution with good generalization performance; such implicit bias is often characterized in terms of the sharpness of the minima. Kleinberg et al. (2018) connected this bias with the smoothing effect of SGD which eliminates sharp local minima by the convolution using the stochastic gradient noise. We follow this line of research and study the commonly-used averaged SGD algorithm, which has been empirically observed in Izmailov et al. (2018) to prefer a flat minimum and therefore achieves better generalization. We prove that in certain problem settings, averaged SGD can efficiently optimize the smoothed objective which avoids sharp local minima. In experiments, we verify our theory and show that parameter averaging with an appropriate step size indeed leads to significant improvement in the performance of SGD.

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。

↑