Abstract
The approximate solution of quasi-linear elliptic boundary value problems by linear programming is considered with emphasis on computational aspects. The solution is obtained as a linear combination of m selected functions, with coefficients determined so as to minimize a weighted sum of the maximum error in the differential equation and on the boundary. A realistic error bound is obtained together with the solution for any value of m, and convergence as $m \to \infty $ is shown under certain conditions. A variety of linear and quasi-linear two-dimensional problems are solved and the numerical results, including error bounds, are presented and discussed.