For half-plane Dirichlet data equal to ReLU^alpha, the harmonic solution lies in the ReLU^alpha-Barron space for 0<alpha<1, fails to lie in it for integer alpha, and admits approximations whose Barron norm grows only logarithmically for integer alpha.
Convergence Rate Analysis for Deep Ritz Method
1 Pith paper cite this work. Polarity classification is still indexing.
abstract
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.
fields
math.AP 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Solving the Poisson Equation with Dirichlet data by shallow ReLU$^\alpha$-networks: A regularity and approximation perspective
For half-plane Dirichlet data equal to ReLU^alpha, the harmonic solution lies in the ReLU^alpha-Barron space for 0<alpha<1, fails to lie in it for integer alpha, and admits approximations whose Barron norm grows only logarithmically for integer alpha.