arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2005.14426cs.LGmath.OCstat.ML
收藏

单神经元梯度下降的不可知学习

Agnostic Learning of a Single Neuron with Gradient Descent

University of California, Los Angeles(加州大学洛杉矶分校)

Spencer Frei, Yuan Cao, Quanquan Gu

摘要:

我们考虑学习最佳拟合的单个神经元的问题,通过使用梯度下降来最小化由一组i.i.d.引起的经验风险,通过在一些未知的联合分布$\mathcal{D}$上的期望平方损失$\mathbb{E}_{(x,y)\sim \mathcal{D}}[(σ(w^\top x)-y)^2]$来测量。样本$S\sim \mathcal{D}^n$。激活函数$σ$是一个任意的Lipschitz非减函数,使得优化问题在一般情况下是非凸和非光滑的,并涵盖了广义线性模型设置下典型的神经网络激活函数和逆连接函数。在不可知的PAC学习设置,其中没有对标签$y$和输入$x$之间的关系的假设,如果最佳的人口风险是$\mathsf{OPT}$,我们表明,梯度下降实现人口风险$O(\mathsf{OPT})+ε$在多项式时间和样本复杂度时,$σ$严格增加。对于ReLU激活,我们的人群风险保证是$O(\mathsf{OPT}^{1/2})+ε$。当标签采取的形式$y = σ(v^\top x) + ξ$为零均值亚高斯噪声$ξ$,我们表明,人口风险保证梯度下降提高到$\mathsf{OPT} + ε$。我们的样本复杂性和运行时保证(几乎)维度独立,当$σ$严格增加时,不需要超出有界性的分布假设。对于ReLU,我们在输入的边际分布的非简并假设下显示了相同的结果。

英文摘要:

We consider the problem of learning the best-fitting single neuron as measured by the expected square loss $\mathbb{E}_{(x,y)\sim \mathcal{D}}[(σ(w^\top x)-y)^2]$ over some unknown joint distribution $\mathcal{D}$ by using gradient descent to minimize the empirical risk induced by a set of i.i.d. samples $S\sim \mathcal{D}^n$. The activation function $σ$ is an arbitrary Lipschitz and non-decreasing function, making the optimization problem nonconvex and nonsmooth in general, and covers typical neural network activation functions and inverse link functions in the generalized linear model setting. In the agnostic PAC learning setting, where no assumption on the relationship between the labels $y$ and the input $x$ is made, if the optimal population risk is $\mathsf{OPT}$, we show that gradient descent achieves population risk $O(\mathsf{OPT})+ε$ in polynomial time and sample complexity when $σ$ is strictly increasing. For the ReLU activation, our population risk guarantee is $O(\mathsf{OPT}^{1/2})+ε$. When labels take the form $y = σ(v^\top x) + ξ$ for zero-mean sub-Gaussian noise $ξ$, we show that the population risk guarantees for gradient descent improve to $\mathsf{OPT} + ε$. Our sample complexity and runtime guarantees are (almost) dimension independent, and when $σ$ is strictly increasing, require no distributional assumptions beyond boundedness. For ReLU, we show the same results under a nondegeneracy assumption for the marginal distribution of the input.

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。

↑