arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2103.13330math.NAcs.NA
收藏

Deep Ritz方法的收敛速度分析

Convergence Rate Analysis for Deep Ritz Method

Chenguang Duan, Yuling Jiao, Yanming Lai, Xiliang Lu, Zhijian Yang

摘要:

使用深度神经网络来解决偏微分方程最近引起了很多关注。然而,深度学习方法的成功原因远远落后于它的经验成功。本文对求解二阶椭圆型方程的Deep Ritz方法(DRM)\cite{wan11}进行了严格的数值分析.我们使用具有$\mathrm{ReLU}^2$激活函数的深度网络建立DRM的$H^1$范数的第一非渐近收敛速度。除了提供DRM的理论依据外,我们的研究还揭示了如何设置深度和宽度的超参数,以实现所需的收敛速度方面的训练样本的数量。在技术上,我们得到了深度$\mathrm{ReLU}^2$网络在$H^1$范数下的逼近误差的界以及梯度范数和$\mathrm{ReLU}^2$网络的非Lipschitz合成的Rademacher复杂度的界,这两个界都是独立的.

英文摘要:

Using deep neural networks to solve PDEs has attracted a lot of attentions recently. However, why the deep learning method works is falling far behind its empirical success. In this paper, we provide a rigorous numerical analysis on deep Ritz method (DRM) \cite{wan11} for second order elliptic equations with Neumann boundary conditions. We establish the first nonasymptotic convergence rate in $H^1$ norm for DRM using deep networks with $\mathrm{ReLU}^2$ activation functions. In addition to providing a theoretical justification of DRM, our study also shed light on how to set the hyper-parameter of depth and width to achieve the desired convergence rate in terms of number of training samples. Technically, we derive bounds on the approximation error of deep $\mathrm{ReLU}^2$ network in $H^1$ norm and on the Rademacher complexity of the non-Lipschitz composition of gradient norm and $\mathrm{ReLU}^2$ network, both of which are of independent interest.

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。

↑