It is widely known that the error analysis for deep learning involves approximation, statistical, and optimization errors. However, it is challenging to combine them together due to overparameterization. In this paper, we address this gap by providing a comprehensive error analysis of the Deep Ritz Method (DRM). Specifically, we investigate a foundational question in the theoretical analysis of DRM under the overparameterized regime: given a target precision level, how can one determine the appropriate number of training samples, the key architectural parameters of the neural networks, the step size for the projected gradient descent optimization procedure, and the requisite number of iterations, such that the output of the gradient descent process closely approximates the true solution of the underlying partial differential equation to the specified precision?

This paper presents a newly developed high-order discontinuous Galerkin time-domain (DGTD) method for solving Maxwell's equations in linear dispersive media with UPML boundary treatment. A unified formulation is derived for linear dispersive media of Debye type and the artificial material in the UPML regions with the help of auxiliary differential equations. The DGTD employs finite-element-type meshes, and uses piecewise high-order polynomials for spatial discretization and Runge-Kutta method for time integrations. Arbitrary high-order accuracy can be obtained for scattering of various objects in dispersive media. After validating the numerical convergence of the DGTD method together with the second-order Yee's scheme, we apply this new method to the ground-penetrating radar for the detection of buried objects in a lossy half space.