1: \begin{abstract}

2:   We introduce a deep learning-based framework for weakly enforcing

3:   boundary conditions in the numerical approximation of partial

4:   differential equations. Building on existing physics-informed neural

5:   network and deep Ritz methods, we propose the Deep Uzawa algorithm,

6:   which incorporates Lagrange multipliers to handle boundary

7:   conditions effectively. This modification requires only a minor

8:   computational adjustment but ensures enhanced convergence properties

9:   and provably accurate enforcement of boundary conditions, even for

10:   singularly perturbed problems.

11: 

12:   We provide a comprehensive mathematical analysis demonstrating the

13:   convergence of the scheme and validate the effectiveness of the Deep

14:   Uzawa algorithm through numerical experiments, including

15:   high-dimensional, singularly perturbed problems and those posed over non-convex domains.

16: \end{abstract}

17: