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 2407.07556math.OCcs.NAmath.NA
收藏

半流中的小批量下降

Mini-batch descent in semiflows

Alberto Domínguez Corella, Martín Hernández

AI总结:

本文针对半流场景,提出连续版小批量梯度下降方法与随机最小化运动格式,证明二者可逼近原始半流轨迹,通过多场景应用与数值算例验证了方法的有效性与通用性。

摘要:

本文研究了小批量梯度下降在半流(梯度流)中的应用。给定损失函数(势函数),我们通过随时间随机选取子损失函数,定义分段流,提出了连续版本的小批量梯度下降方法。我们证明,在生成半流的势函数满足适当假设的条件下,小批量下降流的轨迹平均而言能紧密逼近原始半流的轨迹。此外,我们研究了一种同样能逼近全损失函数半流的随机最小化运动格式。我们通过约束优化、稀疏反演、区域分解等多种问题展示了该方法的通用性,最后通过若干数值算例验证了所得结论。

英文摘要:

This paper investigates the application of mini-batch gradient descent to semiflows (gradient flows). Given a loss function (potential), we introduce a continuous version of mini-batch gradient descent by randomly selecting sub-loss functions over time, defining a piecewise flow. We prove that, under suitable assumptions on the potential generating the semiflow, the \textit{mini-batch descent flow} trajectory closely approximates the original semiflow trajectory on average. In addition, we study a randomized minimizing movement scheme that also approximates the semiflow of the full loss function. We illustrate the versatility of this approach across various problems, including constrained optimization, sparse inversion, and domain decomposition. Finally, we validate our results with several numerical examples.

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。

↑