1: \begin{abstract}

2: Entropy-based ($\M_N$) moment closures for kinetic equations are defined by a

3: constrained optimization problem that must be solved at every point in a

4: space-time mesh, making it important to solve these optimization problems

5: accurately and efficiently. We present a complete and practical numerical

6: algorithm for solving the dual problem in one-dimensional, slab geometries. The

7: closure is only well-defined on the set of moments that are realizable from a

8: positive underlying distribution, and as the boundary of the realizable set is

9: approached, the dual problem becomes increasingly difficult to solve due to

10: ill-conditioning of the Hessian matrix.  To improve the condition number of the

11: Hessian, we advocate the use of a change of polynomial basis, defined using a

12: Cholesky factorization of the Hessian, that permits solution of problems nearer

13: to the boundary of the realizable set. We also advocate a fixed quadrature

14: scheme, rather than adaptive quadrature, since the latter introduces unnecessary

15: expense and changes the computationally realizable set as the quadrature

16: changes. For very ill-conditioned problems, we use regularization to make the

17: optimization algorithm robust. We design a manufactured solution and demonstrate

18: that the adaptive-basis optimization algorithm reduces the need for

19: regularization. This is important since we also show that regularization slows,

20: and even stalls, convergence of the numerical simulation when refining the

21: space-time mesh. We also simulate two well-known benchmark problems. There we

22: find that our adaptive-basis, fixed-quadrature algorithm uses less

23: regularization than alternatives, although differences in the resulting numerical

24: simulations are more sensitive to the regularization strategy than to the choice

25: of basis.

26: 

27: 

28: 

29: 

30: 

31: \end{abstract}

32: