Abstract
Numerical solutions of nonlinear programming problems by the penalty function method are, in general, approximate solutions. An error in these solutions consists of two parts: one is the error between the approximate solution and the true extremal solution of the penalty function, and the other is the error between its true extremal solution and the true solution of the nonlinear programming problem. Urabe's theorem can give the error bound for the former error. In this paper, the error bound theorem for the latter error is given and a numerical technique and a numerical example are given.
