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 1609.08397stat.MLcs.LG
收藏

基于稳定性的优化算法的推广误差界

Generalization Error Bounds for Optimization Algorithms via Stability

School of Mathematical Sciences, Peking University(北京大学数学科学学院);Beijing Jiaotong University(北京交通大学);Microsoft Research(微软研究院);Academy of Mathematics and Systems Science, Chinese Academy of Sciences(中国科学院数学与系统科学研究院)

Qi Meng, Yue Wang, Wei Chen, Taifeng Wang, Zhi-Ming Ma, Tie-Yan Liu

摘要:

许多机器学习任务可以被公式化为正则化经验风险最小化(R-ERM),并通过诸如梯度下降(GD)、随机梯度下降(SGD)和随机方差缩减(SVRG)等优化算法来解决。对这些优化算法的传统分析集中在训练过程中的收敛速度,然而,机器学习社区的人们可能更关心学习模型对未知测试数据的泛化性能。本文以稳定性为工具,对这个问题进行了研究。特别是,我们分解的推广误差的R-ERM,并推导出其上界的凸和非凸的情况下。在凸的情况下,我们证明了推广误差可以有界的收敛速度的优化算法和R-ERM过程的稳定性,无论是在期望(在$\mathcal{O}((1/n)+\mathbb{E}ρ(T))$的顺序,其中$ρ(T)$是收敛误差和$T$是迭代次数)和高概率(在$\mathcal{O}\left(\frac{\log{1/δ}}{\sqrt{n}}+ρ(T)\right)$的顺序与概率$1-δ$)。对于非凸情形,我们也可以得到类似的期望推广误差界。我们的定理表明:1)随着训练过程的进行,所有优化算法的泛化误差都在减小; 2)相对而言,SVRG比GD和SGD具有更好的泛化能力。我们对凸问题和非凸问题进行了实验,实验结果验证了我们的理论研究结果。

英文摘要:

Many machine learning tasks can be formulated as Regularized Empirical Risk Minimization (R-ERM), and solved by optimization algorithms such as gradient descent (GD), stochastic gradient descent (SGD), and stochastic variance reduction (SVRG). Conventional analysis on these optimization algorithms focuses on their convergence rates during the training process, however, people in the machine learning community may care more about the generalization performance of the learned model on unseen test data. In this paper, we investigate on this issue, by using stability as a tool. In particular, we decompose the generalization error for R-ERM, and derive its upper bound for both convex and non-convex cases. In convex cases, we prove that the generalization error can be bounded by the convergence rate of the optimization algorithm and the stability of the R-ERM process, both in expectation (in the order of $\mathcal{O}((1/n)+\mathbb{E}ρ(T))$, where $ρ(T)$ is the convergence error and $T$ is the number of iterations) and in high probability (in the order of $\mathcal{O}\left(\frac{\log{1/δ}}{\sqrt{n}}+ρ(T)\right)$ with probability $1-δ$). For non-convex cases, we can also obtain a similar expected generalization error bound. Our theorems indicate that 1) along with the training process, the generalization error will decrease for all the optimization algorithms under our investigation; 2) Comparatively speaking, SVRG has better generalization ability than GD and SGD. We have conducted experiments on both convex and non-convex problems, and the experimental results verify our theoretical findings.

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。

↑