1: \begin{abstract}

2: Many partial differential equations (PDEs) such as Navier--Stokes equations

3: in fluid mechanics, 

4: inelastic deformation in solids, and transient parabolic and hyperbolic 

5: equations do not have an exact, primal variational structure. Recently, a variational principle based on

6: the dual (Lagrange multiplier) field was proposed.

7: The essential idea in this approach is to treat the given

8: PDE as constraints, and to invoke an arbitrarily chosen 

9: auxiliary potential with strong convexity properties to be optimized.  

10: This leads to requiring a convex dual functional to be minimized subject to 

11: Dirichlet boundary conditions on dual variables, with the guarantee that even 

12: PDEs that do not possess a variational structure in primal form can be solved 

13: via a variational principle. The vanishing of the

14: first variation of the dual functional is, up 

15: to Dirichlet boundary conditions on dual fields, the weak form of the primal PDE problem with the dual-to-primal change of variables incorporated. 

16: We derive the dual weak form for the linear, one-dimensional,  

17: transient convection-diffusion equation. A Galerkin

18: discretization is used to obtain the discrete equations, with

19: the trial and test functions chosen as linear 

20: combination of either RePU activation functions (shallow 

21: neural network) or B-spline basis functions; the corresponding stiffness matrix is symmetric. 

22: For transient problems, a 

23: space-time Galerkin implementation is used with tensor-product

24: B-splines as approximating functions. 

25: Numerical results are presented for 

26: the steady-state and transient convection-diffusion equation, and transient heat conduction. The proposed method

27: delivers sound accuracy for ODEs and PDEs and 

28: rates of convergence are established in the $L^2$ norm and $H^1$ seminorm for the steady-state 

29: convection-diffusion problem.

30: \end{abstract}

31: