arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1802.06175cs.LG
收藏

另一种观点:SGD何时跳出局部极小值?

An Alternative View: When Does SGD Escape Local Minima?

Cornell University(康奈尔大学);Princeton University(普林斯顿大学)

Robert Kleinberg, Yuanzhi Li, Yang Yuan

摘要:

随机梯度下降(SGD)在机器学习中得到了广泛的应用。虽然通常被视为快速但不准确的梯度下降(GD)版本,但它总是比现代神经网络的GD找到更好的解决方案。为了理解这种现象,我们采取了另一种观点,即SGD正在研究损失函数的卷积(因此平滑)版本。我们证明了,即使函数$f$有许多坏的局部极小点或鞍点,只要对每个点$x$,其邻域梯度的加权平均值相对于期望解$x^*$是一个凸点,SGD将以恒定的概率接近并停留在$x^*$附近.更具体地说,SGD不会卡在具有小直径的“尖锐”局部最小值处,只要这些区域的邻域包含足够的梯度信息。邻域大小由步长和梯度噪声控制。我们的结果确定了一组函数,SGD可证明的作品,这是远远大于凸函数的集合。经验上,我们观察到神经网络的损失面局部具有良好的一点凸性,因此我们的定理有助于解释为什么SGD对神经网络如此有效。

英文摘要:

Stochastic gradient descent (SGD) is widely used in machine learning. Although being commonly viewed as a fast but not accurate version of gradient descent (GD), it always finds better solutions than GD for modern neural networks. In order to understand this phenomenon, we take an alternative view that SGD is working on the convolved (thus smoothed) version of the loss function. We show that, even if the function $f$ has many bad local minima or saddle points, as long as for every point $x$, the weighted average of the gradients of its neighborhoods is one point convex with respect to the desired solution $x^*$, SGD will get close to, and then stay around $x^*$ with constant probability. More specifically, SGD will not get stuck at "sharp" local minima with small diameters, as long as the neighborhoods of these regions contain enough gradient information. The neighborhood size is controlled by step size and gradient noise. Our result identifies a set of functions that SGD provably works, which is much larger than the set of convex functions. Empirically, we observe that the loss surface of neural networks enjoys nice one point convexity properties locally, therefore our theorem helps explain why SGD works so well for neural networks.

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。

↑